Internal reference and demo site
EXECUTIVE ASSISTANT
This is the Executive Assistant that runs Lennart Hector's ERP and AI consulting business — a single-operator system built inside Claude Code that handles sales, delivery, finance, people, marketing, and strategy across three legal entities.
Three entities:
Business mix: roughly 90% ERP consulting, 10% AI/cloud/innovation, with the innovation share growing.
Active clients include MeisterWerke, Trojan Technologies, Tsurumi, W. Neudorff, Containex, Calanbau, OTT Hydromet, IMA Schelling, EMG Automation, AMS Technologies, ERMCO-ECI, and others across industrial manufacturing.
Current Q2 priorities: Redeploy three consultants coming off a large engagement ending in June. Close two H2W Labs PoCs. Land the first AWS reference. Drive Moco time tracking adoption past 80%. All tracked in goals.md and scored weekly.
Most of the work of running a consulting business is repetition at scale. Prep a meeting. Follow up on a quote. Score a deal. Draft a reply. Generate a dashboard. Chase an invoice. Prepare a 1:1. Write a status report. Publish a LinkedIn post. Review a PM's draft. Refresh a customer profile.
The traditional answer is to hire more people. The answer this project makes is different: formalize every repeating workflow as a reusable skill, wire up the tools the skill needs, and run the whole thing through one agentic interface. The result is a single operator running at the throughput of a small team.
This site documents every piece: 80+ formalized skills, 7 specialized subagents, 6 behavioral rules, 36+ context files, 35 integration scripts, 8 MCP servers, and a growing library of hand-written workflow playbooks. Everything is plain text in a git repository. You can read the source of every skill, every rule, every context file.
The order is intentional: counts, then structure, then the catalog, then the proof it's alive.
Click any orchestrator to expand its full execution tree. SkillAgentSystemFile
Which data sources each skill connects to. Hover a column to highlight.
| Skill | Close39 | Moco25 | M36512 | Qonto8 | Jira17 | Hedy20 | Apollo7 | Slack2 |
|---|---|---|---|---|---|---|---|---|
| Data | ||||||||
| context-refresh | ||||||||
| weekly-snapshot | ||||||||
| Delivery | ||||||||
| account-status-report | ||||||||
| labs-demo-delivery | ||||||||
| project-status-reporting | ||||||||
| qbr-client-review | ||||||||
| inbox-drafts | ||||||||
| Finance | ||||||||
| cash-flow-forecast | ||||||||
| cfo-analysis | ||||||||
| intercompany-billing | ||||||||
| invoice-aging | ||||||||
| Hedy/Automation | ||||||||
| hedy-session-tagger | ||||||||
| Marketing | ||||||||
| ab-testing | ||||||||
| cmo-analysis | ||||||||
| content-analytics | ||||||||
| newsletter-assembler | ||||||||
| Meetings | ||||||||
| account-management-meeting | ||||||||
| daily-calendar-scan | ||||||||
| meeting-briefing-notes | ||||||||
| meeting-workflow | ||||||||
| one-on-one-prep | ||||||||
| Meta | ||||||||
| skill-pattern-detector | ||||||||
| Operations | ||||||||
| action-item-tracker | ||||||||
| moco-crm-sync | ||||||||
| operations-health | ||||||||
| utilization-report | ||||||||
| workforce-planning | ||||||||
| People | ||||||||
| chro-analysis | ||||||||
| evaluation-cycle | ||||||||
| people-prep | ||||||||
| relationship-pulse | ||||||||
| Product | ||||||||
| labs-launch-strategy | ||||||||
| labs-product-management | ||||||||
| Reporting | ||||||||
| monday-briefing | ||||||||
| top-management-report | ||||||||
| weekly-review | ||||||||
| Revenue | ||||||||
| cro-analysis | ||||||||
| Sales | ||||||||
| customer-intelligence | ||||||||
| deal-pulse | ||||||||
| email-sequence-design | ||||||||
| h2w-labs-playbook | ||||||||
| pipeline-velocity | ||||||||
| prospecting-engine | ||||||||
| sales-discovery-debrief | ||||||||
| sales-intelligence | ||||||||
| Session | ||||||||
| closeout | ||||||||
| Strategy | ||||||||
| competitive-intelligence | ||||||||
| cso-analysis | ||||||||
| market-sizing | ||||||||
Every recurring task the EA runs, mapped to when it fires. Click any pill for details.
"Aggregate open action items across Hedy meetings, Jira, M365 Tasks, and Close CRM. Surface overdue items, group by owner, flag items ope...
"What's overdue", "open action items", "who owes what"
"Quick business alert scan using local files only (no API calls). Use at session start or when Lennart says check alerts, what needs atte...
"Check alerts", "any red flags?", "what needs attention?"
Content production orchestrator. Checks queue health, drafts LinkedIn posts and blog articles in batch, manages the 3-day approved buffer...
"Fill the queue", "draft posts", "queue status"
Cross-functional daily business scan. Use when Lennart says what is on my calendar, anything new, what should I know, daily scan, or at s...
"What's on my calendar", "daily scan", "calendar check"
Sales execution loop that monitors email and calendar for deal signals, flags overdue follow-ups, and surfaces deals needing action. Use...
"Deal pulse", "check my deals", "follow-up check"
"Check for unsent email drafts and remind about stale ones. Use when Lennart says check drafts, pending drafts, draft status, unsent emai...
"Check drafts", "pending drafts", "unsent emails"
"Scan inbox for emails needing replies, draft responses for all of them. Runs as part of daily-calendar-scan or standalone. Use when Lenn...
"Inbox drafts", "draft my replies", "what needs a reply"
"Log daily time entries to Moco based on calendar events and manual input. Use when Lennart says log time, Zeiterfassung, log my hours, t...
"Log time", "Zeiterfassung", "log my hours". Interactive version with approval gate. Closeout skill handles autonomous posting.
Full meeting lifecycle -- preparation, notes processing, and follow-up drafting. Use when Lennart mentions meeting prep, meeting notes, m...
"Meeting prep", "process meeting notes"
Run the full Monday morning report sequence in one command. Chains context-refresh, c-suite-dashboard, monday-briefing, and weekly-snapsh...
"Monday sequence", "run the Monday reports"
Process a sales discovery meeting into actionable CRM updates, deal qualification, and follow-up emails. Pulls the Hedy transcript, extra...
"Debrief the call", "process the meeting"
"DEPRECATED -- use /closeout instead. Redirects to the closeout skill."
A/B testing for email sequences and content. Design hypotheses, set up email split tests in Close CRM, design content experiments for LinkedIn and blog, analyze results. Use when Lennart says A/B test, split test, tes...
Triggers: "A/B test", "split test", "test subject line"
Standardized account management meeting framework. Generates internal prep brief (PDF) and customer-facing deck (PPTX) for monthly account reviews. Use when Lennart says account review, account management meeting, pre...
Triggers: "Account meeting prep [client]", "customer review for [client]"
Triggers: "Account status report", "customer report for [client]"
"Aggregate open action items across Hedy meetings, Jira, M365 Tasks, and Close CRM. Surface overdue items, group by owner, flag items open 7+ days. Use when Lennart asks what's overdue, open action items, who owes wha...
Triggers: "What's overdue", "open action items", "who owes what"
"Quick business alert scan using local files only (no API calls). Use at session start or when Lennart says check alerts, what needs attention, any red flags, or business pulse."
Triggers: "Check alerts", "any red flags?", "what needs attention?"
SEO-targeted blog article production with bilingual output, meta descriptions, and keyword targeting. Use when Lennart says write a blog post, draft an article, blog about [topic], SEO article, or create blog content.
Triggers: "Write blog article", "blog on [topic]"
Run all C-level analysis dashboards and produce a combined executive report. Chains CFO, CRO, CHRO, CSO, and CMO weekly dashboards into individual PDFs plus one combined C-suite dashboard PDF. Use when Lennart says C-...
Triggers: "C-suite dashboard", "executive dashboard"
Rolling 13-week cash flow forecast combining Qonto bank data, Moco invoices, Close CRM pipeline, and known fixed costs. Use when Lennart asks about cash flow, cash runway, when we run out of money, upcoming payments,...
Triggers: "Cash flow forecast", "cash runway"
CFO-level financial analysis with weekly dashboard, scenario modeling, deal economics, and investment decisions. Use when Lennart asks about margins, profitability, pricing, deal economics, revenue scenarios, cost ana...
Triggers: "Deal economics", "revenue scenarios", "CFO dashboard"
"Check Claude usage limits via Playwright and cache the result. Use at session start, before heavy work, or when approaching limits. Also triggers auto-switch to OpenRouter if limits are critical."
CHRO-level people strategy with weekly dashboard, retention risk assessment, succession planning, comp benchmarking, and org design. Use when Lennart asks about team health, retention risks, bench cost, redeployment,...
Triggers: "Team health", "retention risks", "CHRO dashboard"
Create, update, or deploy password-protected client microsites. Use when Lennart says create a client website, build a site for [company], positioning website, proposal site, workshop results site, QBR site, update th...
"Autonomous end-of-day closeout: git commit, Moco time logging, CRM sync, daily log, session briefing, self-validation. Use when Lennart says closeout, wrap up, end session, done for now, save state. Replaces session-...
Triggers: "Wrap up", "end session", "closeout", "done for now", "save state". Modes: --quick (git+briefing+time-split), --dry-run (preview). Replaces session-closeout.
CMO-level marketing analysis with weekly dashboard, content ROI, channel effectiveness, and brand health. Use when Lennart asks about marketing performance, content output, lead attribution, brand visibility, content...
Triggers: "Marketing performance", "content output", "CMO dashboard"
Competitive landscape mapping, deal battle cards, win/loss pattern analysis, and comparison content generation. Use when Lennart says competitor analysis, battle card, competitive matrix, win/loss analysis, comparison...
Triggers: "Competitor analysis", "battle card", "win/loss analysis"
Content performance tracking, weekly and monthly analytics reports, and pipeline attribution. Use when Lennart says content analytics, content performance, what's working, weekly content report, or attribution report.
Triggers: "Content analytics", "weekly content report"
Content production orchestrator. Checks queue health, drafts LinkedIn posts and blog articles in batch, manages the 3-day approved buffer, and stages content for review. Use when Lennart says content engine, fill the...
Triggers: "Fill the queue", "draft posts", "queue status"
Batch content review workflow. Presents drafted LinkedIn posts and blog articles for Lennart's review, tracks edits for voice calibration, and moves approved content to the publishing queue. Use when Lennart says revi...
Triggers: "Review content", "approve posts", "content review"
Refresh context files with live data from Moco, Close CRM, and Qonto. Run weekly (before Monday reports) or on demand to keep context files current. Use when Lennart says refresh context, update context, sync data, pu...
Triggers: "Refresh context", "pull latest numbers"
[triggers: improve copy, review this draft, make this more compelling, copy review, sharpen the messaging, proposal copy, landing page copy]
Triggers: "Improve copy", "copy review", "sharpen messaging"
[triggers: edit this, review this copy, proofread, quality check, polish this draft, improve this text, copy edit, seven sweep]
Triggers: "Edit this", "proofread", "quality check", "seven sweep"
Run CRM and Moco data quality validation. Use when Lennart says validate CRM, check data quality, run CRM audit, or after any Close/Moco write operation.
CRO-level revenue analysis with weekly dashboard, deal intelligence, revenue forecasting, and sales performance. Use when Lennart asks about pipeline health, revenue forecast, sales velocity, deal prioritization, conv...
Triggers: "Pipeline health", "revenue forecast", "CRO dashboard"
CSO-level strategic analysis with weekly dashboard, competitive intelligence, strategic decision support, and market positioning review. Use when Lennart asks about strategic bets, risk register, competitive moves, ma...
Triggers: "Strategic bets", "competitive intel", "CSO dashboard"
Build deep, comprehensive customer profiles by pulling from all systems (Close CRM, Jira, Moco, Hedy, Apollo.io, web). Use when Lennart asks to understand a customer deeply, build a customer profile, analyze customer...
Triggers: "Customer profile", "stakeholder map"
Cross-functional daily business scan. Use when Lennart says what is on my calendar, anything new, what should I know, daily scan, or at session start. Covers all 6 business areas (sales, delivery, people, finance, ope...
Triggers: "What's on my calendar", "daily scan", "calendar check"
"Create a daily log entry summarizing EA project activity and company business events. Use when Lennart says daily log, log today, what happened today, end of day, or automatically via evening cron."
[triggers: dashboard layout, KPI design, improve this report, report design, dashboard design, better visualization, report layout]
Triggers: "Dashboard layout", "KPI design", "report design"
Score all qualified pipeline opportunities (Interested but need time through Verge of closing) on a 0-100 health scale and tier them Green/Yellow/Red/Stalled. Use when Lennart asks to score deals, check deal health, f...
Sales execution loop that monitors email and calendar for deal signals, flags overdue follow-ups, and surfaces deals needing action. Use when Lennart says deal pulse, check my deals, any replies, follow-up check, what...
Triggers: "Deal pulse", "check my deals", "follow-up check"
"Check for unsent email drafts and remind about stale ones. Use when Lennart says check drafts, pending drafts, draft status, unsent emails, what drafts are waiting, or did I send that email."
Triggers: "Check drafts", "pending drafts", "unsent emails"
CRM contact enrichment procedures, custom field IDs, and email drafting checklists. Auto-loads when creating/updating contacts, drafting emails, or doing CRM enrichment work.
Design and build email sequences for Close CRM. Sales follow-up, nurture, re-engagement, workshop invitations, Labs product intros, and multi-touch campaigns. Use when Lennart says build a sequence, design email caden...
Triggers: "Build a sequence", "design email cadence", "drip campaign"
"Run the full employee evaluation cycle: prep, agenda, and post-meeting record update. Use when Lennart says evaluation for [name], run evaluation cycle, performance review for [name], or after completing an evaluatio...
Triggers: "Evaluation for [name]", "performance review for [name]"
Run the H2W Labs sales conversion playbook for a specific deal. Generates Opportunity Briefs, demo prep, proposals, and follow-up cadences based on deal stage. Use when Lennart asks to move a Labs deal forward, genera...
Triggers: "Run the playbook for [company]"
Auto-tag Hedy sessions to topics and update the local session index. Called by daily-calendar-scan at end of each run. No manual involvement. Use when sessions need tagging, index needs updating, or as part of daily/c...
Triggers: Auto-tags sessions to topics, updates index. Called by daily-calendar-scan and context-refresh.
"Scan inbox for emails needing replies, draft responses for all of them. Runs as part of daily-calendar-scan or standalone. Use when Lennart asks to check inbox, draft replies, process emails, clear inbox, what needs...
Triggers: "Inbox drafts", "draft my replies", "what needs a reply"
"Track intercompany billing between BLC UG and H2W GmbH. Reconcile invoices against bank transfers, flag missing payments, detect entity confusion. Use when Lennart asks about intercompany status, BLC to H2W transfers...
Triggers: "Intercompany status", "BLC to H2W transfers"
Invoice aging report from Moco and Qonto. Shows outstanding receivables by customer and age bucket, overdue invoices, collection risk, and payment patterns. Use when Lennart asks about outstanding invoices, overdue pa...
Triggers: "Outstanding invoices", "who owes us"
"Track issues, tasks, and topics with internal and external people and companies. Use when Lennart says add issue, track this, open issues, issue list, prep for meeting with X, issue dashboard, update issue, close iss...
Triggers: "Add issue", "track this", "open issues", "prep for [company]", "issue dashboard"
"Prepare and structure H2W Labs demo and PoC deliveries. Generates demo scripts, environment checklists, success criteria, and post-demo follow-up. Use when Lennart asks to prep a demo, build a demo script, PoC delive...
Triggers: "Prep demo for [company]", "PoC delivery plan"
Build and execute go-to-market launch plans for H2W Labs products. Positioning, messaging, reference cases, launch sequences, channel coordination. Use when Lennart says launch strategy, go-to-market, GTM plan, produc...
Triggers: "Launch strategy", "GTM plan", "product positioning"
Manage H2W Labs product roadmaps, PoC pipeline, customer feedback synthesis, product briefs, and conversion funnel analysis. Use when Lennart asks about Labs product priorities, feature backlog, PoC status, product fe...
Triggers: "Labs roadmap", "PoC pipeline", "product brief"
"Log daily time entries to Moco based on calendar events and manual input. Use when Lennart says log time, Zeiterfassung, log my hours, time tracking, update Moco time, or as part of session-closeout."
Triggers: "Log time", "Zeiterfassung", "log my hours". Interactive version with approval gate. Closeout skill handles autonomous posting.
Size markets, analyze segments, and assess expansion opportunities for ERP consulting and AI products. Use when Lennart says market size, TAM SAM SOM, market analysis, addressable market, market opportunity, how big i...
Triggers: "Market size", "TAM SAM SOM", "addressable market"
Turn project experiences, customer conversations, and operational insights into content across multiple formats and channels. Use when Lennart asks to write a LinkedIn post, create a case study, draft an article, gene...
Triggers: "Write LinkedIn post", "create case study"
Parallel meeting prep engine. Spawns 4 subagents (Close CRM, Jira, Moco, vault+web) simultaneously, merges outputs, computes relationship health score, and produces an extended briefing note as markdown and one-page P...
Triggers: "Prep for [client]", "research [company]", "meeting prep"
Full meeting lifecycle -- preparation, notes processing, and follow-up drafting. Use when Lennart mentions meeting prep, meeting notes, meeting transcript, follow-up emails, post-meeting actions, or asks to prepare fo...
Triggers: "Meeting prep", "process meeting notes"
Cross-check Moco projects/deals against Close CRM opportunities. Flag misalignments, missing entries, and value discrepancies.
Triggers: "Sync check", "deal alignment", "moco crm sync"
Generate the Monday morning BD-PMO weekly update call briefing. Pulls from Close CRM, Jira, Hedy, Moco, and Slack to prepare agenda talking points. Use when Lennart says Monday briefing, prep the Monday meeting, BD-PM...
Triggers: "Monday briefing", "prep the Monday meeting"
Run the full Monday morning report sequence in one command. Chains context-refresh, c-suite-dashboard, monday-briefing, and weekly-snapshot in the correct order with consistent data. The c-suite-dashboard replaces the...
Triggers: "Monday sequence", "run the Monday reports"
Monthly bilingual newsletter compilation from published content. Curates top pieces, creates Close CRM email templates and workflows for sending. Use when Lennart says assemble newsletter, monthly newsletter, prepare...
Triggers: "Assemble newsletter", "monthly newsletter"
"Prepare lightweight briefings for recurring 1:1 meetings with team members. Pulls last meeting notes, recent Jira activity, Moco hours, and open action items. Use when Lennart asks to prep for a 1:1 with Uli, Richard...
Triggers: "Prep for 1:1 with [person]", "catch me up on [person]"
Operational health monitoring across utilization, delivery, pipeline, and finance. Use when Lennart asks for an ops dashboard, weekly ops check, KPI review, how the business is running, delivery health, or wants a cro...
Triggers: "Ops dashboard", "KPI review"
"Pull a comprehensive briefing on any team member by combining their people record with fresh Moco, email, Jira, and Hedy data. Use when Lennart says prep for [name], catch me up on [name], what's [name] working on, o...
Triggers: "Prep for [name]", "catch me up on [name]"
Weekly pipeline velocity report from Close CRM. Shows deals by stage, age in stage, stalled deals, conversion rates, weighted revenue, and win/loss analysis. Use when Lennart asks about pipeline velocity, deal flow, s...
Triggers: "Pipeline velocity", "stalled deals"
Generate and edit PowerPoint presentations from templates. Use when user needs to create slides from outlines, modify existing decks, profile custom templates, or validate presentation quality.
Triggers: Engine: outline to PPTX, edit decks, validate, profile templates
Create on-brand PPTX presentations for any topic. Use when user asks to create a presentation, slide deck, pitch deck, or PPTX file.
Triggers: "Create a deck", "pitch deck", "slide deck", "PPTX"
Validate, repair, and audit PowerPoint presentations. Use when polishing a generated deck, checking an existing PPTX for quality issues, or running QA before delivery.
Triggers: QA: audit, repair, validate any PPTX. Text overflow, ghosts, fonts, hyperlinks, ZIP fixes
Scaled content strategy design for programmatic SEO. Evaluates feasibility, designs template-based page strategies, and generates keyword x topic matrices for batch production. Use when Lennart says programmatic SEO,...
Triggers: "Programmatic SEO", "scaled content", "keyword matrix"
Regenerate and deploy the password-gated EA project showcase website at webapp/showcase/. Use when Lennart says regenerate showcase, update showcase, rebuild the showcase site, deploy showcase, project showcase, or sh...
Generate weekly project status reports in client-facing and internal versions. Use when Lennart asks to write a status report, prepare a project update, review a PM's status report draft, or summarize project status f...
Triggers: "Status report for [client]"
Generate consulting proposals and statements of work for ERP consulting and H2W Labs engagements. Use when Lennart asks to draft a proposal, write a SOW, put together an offer, create a quote, or prepare commercial do...
Triggers: "Draft a proposal", "write a SOW"
Build and manage target account lists for systematic outbound prospecting. Uses Apollo.io to find ICP-matching companies, enriches with contact data, and stages outreach in Close CRM. Use when Lennart asks to build a...
Triggers: "Build prospect list", "find target accounts"
Scrape web sources, qualify companies against ICP, enrich via Apollo, and push to Close CRM with follow-up tasks. Use when Lennart says scrape and prospect, intake from URL, prospect from [source], run intake on, pros...
"Prepare quarterly business reviews with key clients. Generates QBR agenda, delivery summary, financials, expansion opportunities, and follow-up actions. Use when Lennart asks to prepare a QBR, quarterly review with a...
Triggers: "Prepare QBR for [client]", "quarterly review with [client]"
> Relationship intelligence dashboard. Monitors interaction frequency with key contacts and alerts when relationships need attention. Use when Lennart says relationship pulse, check my relationships, who needs attenti...
Triggers: "Relationship pulse", "who needs attention", "check my relationships"
Process a sales discovery meeting into actionable CRM updates, deal qualification, and follow-up emails. Pulls the Hedy transcript, extracts qualifying information (pain points, budget, timeline, decision makers, tech...
Triggers: "Debrief the call", "process the meeting"
Sales cockpit that pulls data from Close CRM, cross-references with Jira delivery, enriches with Apollo.io and web research. Use when Lennart asks about the pipeline, deals needing attention, a specific deal or compan...
Triggers: "Pipeline review", "deal intelligence"
Draft outreach messages for sales scenarios across email and LinkedIn. Use when Lennart asks to write an outreach email, draft a LinkedIn message, follow up with a contact, re-engage a dormant lead, respond to an inbo...
Triggers: "Write outreach email", "follow up with..."
SEO auditing for h2w-consulting.com. Site health, content scoring, keyword strategy, and competitor SEO analysis. Use when Lennart says SEO audit, content audit, keyword review, SEO health, competitor SEO, or ranking...
Triggers: "SEO audit", "content audit", "keyword review"
"DEPRECATED -- use /closeout instead. Redirects to the closeout skill."
Optimize the prompts, structure, and efficiency of existing skills and agents. Use when Lennart says optimize skill, improve skill, skill quality, prompt optimization, reduce token usage, skill review, make this skill...
Triggers: "Optimize skill", "skill review", "prompt optimization"
Analyze recent conversations, meetings, and workflows to detect recurring patterns that should become formalized skills. Use when Lennart asks to find skill gaps, review what should be automated, identify missing skil...
Triggers: "What skills am I missing?"
Quarterly strategy review and strategic decision support. Use when Lennart asks for a quarterly review, strategy assessment, market positioning analysis, strategic decision, or wants to update the growth thesis. Conne...
Triggers: "Quarterly review", "strategic assessment"
"Switch between model profiles (Claude direct, OpenRouter free/paid/turbo). Use when hitting rate limits, when Lennart says switch models, use free models, use cheaper models, go faster, or go back to Claude."
"[DEPRECATED] Replaced by c-suite-dashboard. Use c-suite-dashboard for the weekly executive report. The CFO dashboard mode covers all financial analysis from this report. Archived copy in archives/skills/top-managemen...
Triggers: "MD report", "cash position"
Consultant and team utilization reporting from Moco. Shows billable ratios, capacity gaps, bench risk, and resource forecast. Use when Lennart asks about utilization, billable hours, team capacity, who is available, b...
Triggers: "Utilization", "billable hours"
Self-improvement loop for the Obsidian vault. Detects new entities in daily logs and transcripts, backfills missing Dataview inline fields on opportunities, refreshes MOCs, validates wiki-aliases, and proposes structu...
"[triggers: /visualize, /visual-plan, /visual-slides, /diff-review, /plan-review, /project-recap, /fact-check, /share, visualize, diagram, visual, interactive HTML, architecture diagram, data flow, slide deck, visual...
Generate a comprehensive weekly review report for Friday leadership meetings. Pulls from Hedy, Close CRM, Jira, and Slack to synthesize the week into an actionable briefing. Use when Lennart says weekly review, Friday...
Triggers: "Weekly review", "Friday briefing"
Capture weekly business metrics snapshot for trend analysis. Run after Monday reports to save key numbers (cash, pipeline, utilization, delivery, Labs) to reports/internal/snapshots/. Use when Lennart says save the sn...
Triggers: "Save the snapshot", "capture this week's numbers"
Workforce planning, team capacity analysis, and people decisions. Use when Lennart asks about staffing, hiring, team redeployment, skill gaps, bench risk, Neudorff impact on team, who to assign to new projects, or wor...
Triggers: "Staffing", "[[w-neudorff
Prepare agendas, materials, checklists, and follow-up documents for customer workshops. Use when Lennart asks to prepare a workshop, draft a workshop agenda, write up workshop results, create a workshop proposal, or b...
Triggers: "Prepare workshop for [company]"
Click any skill card to see its full body.
A/B testing for email sequences and content. Design hypotheses, set up email split tests in Close CRM, design content experiments for LinkedIn and blog, analyze results. Use when Lennart says A/B test, split test, test this subject line, which version works better, email test, or content experiment.
Triggers: "A/B test", "split test", "test subject line"
Data sources: Close, context files
Designs, executes, and analyzes A/B tests for email sequences (Close CRM) and content (LinkedIn posts, blog articles). Provides structured hypothesis design, sample size guidance calibrated for B2B consulting volumes, and statistical rigor appropriate for small audiences.
This skill DESIGNS tests and ANALYZES results. email-sequence-design builds the sequences. content-engine produces the content variants. content-analytics tracks performance data.
Use this skill when Lennart says:
Also proactively suggest:
Always read before execution:
context/marketing.md -- Content pillars, channel targets, benchmarkscontext/sales-playbook.md -- Email cadences, messaging patternscontext/icp.md -- Buyer personas, industry segmentsRead on demand:
context/win-loss.md -- What messaging resonates and what falls flatcontext/seo-keywords.md -- Keyword performance for blog test designdata/content-queue/published/ -- Published content with performance dataB2B ERP consulting operates with small audiences. Tests must be designed accordingly.
| Channel | Minimum per Variant | Why | Practical Implication |
|---|---|---|---|
| Email (subject line) | 200 contacts | Need detectable open rate difference of 5+ percentage points | Split large sends; may take 2-4 sends to accumulate |
| Email (body/CTA) | 200 contacts | Click and reply rates are lower, need more data | Same send pooling as above |
| Email (send time) | 100 contacts | Comparing same content at different times | Run over 2-4 weeks on the same segment |
...
Standardized account management meeting framework. Generates internal prep brief (PDF) and customer-facing deck (PPTX) for monthly account reviews. Use when Lennart says account review, account management meeting, prep for [customer] review, customer deck, or account meeting for [customer].
Triggers: "Account meeting prep [client]", "customer review for [client]"
Data sources: Close, Moco, Jira, Hedy, M365
Standardized framework for monthly account management meetings with key customers. Produces two outputs:
Use this skill when Lennart says:
| Mode | Command | Output |
|---|---|---|
both (default) | /account-management-meeting [customer] | PDF brief + PPTX deck |
prep | /account-management-meeting [customer] --prep | PDF brief only |
deck | /account-management-meeting [customer] --deck | PPTX deck only |
followup | /account-management-meeting [customer] --followup | Delegates to meeting-workflow skill |
context/clients.mdLaunch these 5 subagents in parallel to gather live data. Pass the customer name to each.
Subagent: close-crm-agent
Task: Pull all data for [customer]:
- Lead record (search by company name)
- All contacts on the lead
- Active opportunities (find_opportunities with lead_id filter)
- Won opportunities (last 12 months)
- Activity history (last 90 days): emails, calls, meetings, notes
- Activity count and frequencySubagent: moco-data-agent
Task: Pull project and financial data for [customer]:
- All active projects (filter by customer name)
- Budget and hours consumed per project
- Team assignments per project
- Recent time entries (last 30 days)
- Invoices (last 6 months): amounts, status, payment dates
- Any active offers/quotesSubagent: jira-agent
Task: Pull delivery data for [customer]:
- cloudId: 0ee5c601-cc7e-4022-ba05-749fee885399
...Triggers: "Account status report", "customer report for [client]"
Data sources: Close, Jira, Moco
Generate a branded PDF account status report for any customer account. Pulls live data from Close CRM, Jira, and Moco, then renders a structured PDF using the report engine.
Use when Lennart asks for an account report, account status, account overview PDF, customer report, or wants to share an account summary with the team responsible for that account.
Run these three subagents in parallel to collect account data:
project = {PROJECT_KEY} ORDER BY updated DESC (find project key from projects/ folder or ask)project = PKL AND summary ~ "{company}" ORDER BY updated DESCprojects/{account}/README.md for cached Moco project dataGET /api/v1/projects?identifier={search} for budgets and teamThe report follows a fixed 7-section structure:
Plus:
import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", ".."))
from reports.account_status_report.generate import generate_account_report
data = {
"account_name": "Company Name",
"date": "YYYY-MM-DD",
"entity": "Contracting entity",
"pmo": "PMO Lead Name",
"status": "Active", # Active / At Risk / Inactive
"icp_priority": "A", # A / B / C
..."Aggregate open action items across Hedy meetings, Jira, M365 Tasks, and Close CRM. Surface overdue items, group by owner, flag items open 7+ days. Use when Lennart asks what's overdue, open action items, who owes what, outstanding tasks, action item review, or wants a cross-system task view."
Triggers: "What's overdue", "open action items", "who owes what"
Data sources: Hedy, Jira, M365, Close, Issues
Aggregate open action items from all systems into a single view. No more hunting across Hedy, Jira, M365 Tasks, and CRM to find what's outstanding.
Use this skill when Lennart says:
GetSessions -- recent meetings (last 2-4 weeks)GetSessionDetails -- get full transcript (transcript or cleaned_transcript), then extract action items and owners from the raw conversationsearchJiraIssuesUsingJql -- overdue issues, issues assigned but not started0ee5c601-cc7e-4022-ba05-749fee885399bash scripts/m365.sh todo-lists -- get task listsbash scripts/m365.sh todo-tasks [list-id] -- get tasks from each listactivity_search -- recent notes and tasksfind_opportunities -- deals with next_action_date in the past (overdue follow-ups)data/issues/ISS-*.md -- all issue filesHedy (main conversation):
GetSessions limit 15 -- last 2-3 weeks of meetingsGetSessionDetails for each session -- get full transcript (transcript or cleaned_transcript)Jira (jira-agent):
searchJiraIssuesUsingJql: duedate < now() AND status != Done ORDER BY duedate ASC -- all overduesearchJiraIssuesUsingJql: status = "In Progress" AND updated < -7d -- stale in-progress itemsM365 Tasks (main conversation, bash):
bash scripts/m365.sh todo-lists -- list all task lists...
"Quick business alert scan using local files only (no API calls). Use at session start or when Lennart says check alerts, what needs attention, any red flags, or business pulse."
Triggers: "Check alerts", "any red flags?", "what needs attention?"
Data sources: Snapshots, context files, data/issues/
Fast business alert scan. Reads local files only (no API calls). Surfaces what needs immediate attention.
Use when Lennart says:
Read these local files. Do NOT make API calls -- this skill must be fast.
reports/internal/snapshots/ (skip README.md)context/pipeline.md (check "Last updated" date + deal data)context/contracts.md (check expiry dates)data/issues/ISS-*.md, read frontmatter only (status, due_date, priority)data/session-briefing.md (data freshness table)data/draft-log.md (check for stale unsent drafts)| Check | Source | RED | YELLOW |
|---|---|---|---|
| H2W GmbH cash | Snapshot | < EUR 80K | < EUR 120K |
| Group cash | Snapshot | < EUR 150K | < EUR 200K |
| Pipeline coverage | Snapshot/pipeline | < 1.0x monthly revenue | < 2.0x |
| Data staleness | "Last updated" in pipeline.md, projects.md, team-assignments.md | > 14 days | > 7 days |
| Deal aging | pipeline.md proposals | Any proposal > 21 days no activity | > 14 days |
| Overdue issues | data/issues/ frontmatter | Any past due_date with status open/in-progress | Due within 3 days |
| Contract expiry | contracts.md | Expiring < 14 days | Expiring < 30 days |
| Neudorff countdown | Calculate from today to June 30, 2026 | < 8 weeks + EUR 0 replacement | < 16 weeks |
| Stale drafts | data/draft-log.md | Any draft past RED threshold for its urgency tier | Any draft past YELLOW threshold |
Compact alert block. RED items first, then YELLOW.
ALERTS ([count] active)
RED [icon-free label]: [metric] -- [suggested action or skill]
RED [label]: [metric] -- [action]
YELLOW [label]: [metric] -- [action]
No alerts: all clear.Examples:
ALERTS (3 active)
RED Cash: H2W EUR 18K (threshold EUR 80K) -- /invoice-aging or /cfo-analysis
RED Pipeline: 0.4x coverage, J.P. Sauer proposal 14 days aging -- /deal-pulse
YELLOW Data: pipeline.md 8 days stale -- /context-refresh...
SEO-targeted blog article production with bilingual output, meta descriptions, and keyword targeting. Use when Lennart says write a blog post, draft an article, blog about [topic], SEO article, or create blog content.
Triggers: "Write blog article", "blog on [topic]"
Data sources: SEO keywords, context files
Produces SEO-targeted blog articles for the H2W website. Each article targets one keyword from the SEO strategy, follows the weekly publishing schedule, and includes all metadata needed for CMS publishing.
Use this skill when Lennart says:
Also triggered by content-engine Mode 2 for scheduled blog production.
Always read before writing:
context/seo-keywords.md -- Target keywords and ranking statuscontext/offerings.md -- Product/service details for accuracycontext/marketing.md -- Content pillars, existing content inventorycontext/icp.md -- Buyer personas (write for these readers)context/strategy.md -- Current strategic focus| Day | Language | Article Type | Focus |
|---|---|---|---|
| Mon | DE | SEO article | Infor LN keyword targeting |
| Tue | EN | Thought leadership | Industry insight, long-form opinion |
| Wed | DE | How-to / practical guide | Step-by-step, actionable |
| Thu | EN | Case study or use case | Real project experience (anonymized) |
| Fri | DE | Industry trend / opinion | Manufacturing industry direction |
| Sat | EN | AI in manufacturing | H2W Labs topics, practical AI |
| Sun | DE | Evergreen / pillar content | Foundational topics that rank long-term |
Every article must include:
Frontmatter (for CMS and queue tracking):
---
title: [Article title -- include target keyword naturally]
meta_description: [150-160 characters, include target keyword, compelling to click]
target_keyword: [Primary keyword from seo-keywords.md]
secondary_keywords: [2-3 related terms to include naturally]
language: EN|DE
article_type: seo|thought-leadership|how-to|case-study|trend|ai-manufacturing|pillar
pillar: [Content pillar from marketing.md]
icp: ICP1|ICP2|Both
author: Lennart Hector
status: drafted
created: YYYY-MM-DD
word_count: [actual count]
---Article body (800-1500 words):
...
Run all C-level analysis dashboards and produce a combined executive report. Chains CFO, CRO, CHRO, CSO, and CMO weekly dashboards into individual PDFs plus one combined C-suite dashboard PDF. Use when Lennart says C-suite dashboard, executive dashboard, full leadership view, run all C-level reports, or as Step 2 in the Monday sequence.
Triggers: "C-suite dashboard", "executive dashboard"
Data sources: All
Runs all five C-level analysis skills in their weekly dashboard mode and produces:
This is the core executive report in the Monday sequence, replacing the former top-management-report.
Use this skill when Lennart says:
Launch all data-gathering agents simultaneously to minimize runtime:
Agent 1: finance-data-agent
Agent 2: moco-data-agent
Agent 3: close-crm-agent
Close REST API (run from main conversation, parallel with agents):
export $(grep CLOSE_API_KEY .env)
ORG_ID="orga_FLXmECEurzttNYuIdoJNHHR5P83kaC3SNvgCjkgZ7RH"
# Activity metrics this week (for CRO + CMO sections)
curl -s -u "$CLOSE_API_KEY:" -X POST -H "Content-Type: application/json" \
-H "x-tz-offset: -4" \
-d '{"type": "overview", "metrics": ["calls.count","emails.sent.count","emails.received.count","meetings.count","opportunities.created.count","opportunities.won.count","opportunities.lost.count"], "relative_range": "this-week"}' \
"https://api.close.com/api/v1/report/activity/"
# Opportunity status transitions this week (for CRO section)
curl -s -u "$CLOSE_API_KEY:" -H "x-tz-offset: -4" \
"https://api.close.com/api/v1/report/statuses/opportunity/$ORG_ID/?date_start=$(date -d '7 days ago' +%Y-%m-%d)&date_end=$(date +%Y-%m-%d)"
# Pipeline funnel (for CRO section)
curl -s -u "$CLOSE_API_KEY:" -X POST -H "Content-Type: application/json" \
-H "x-tz-offset: -4" \
-d '{"pipeline": "pipe_1yf3mun5yjz4y7AjJ08YVb", "type": "created-cohort", "relative_range": "this-quarter"}' \
"https://api.close.com/api/v1/report/funnel/opportunity/totals/"
# Email template stats (for CMO section)
...Rolling 13-week cash flow forecast combining Qonto bank data, Moco invoices, Close CRM pipeline, and known fixed costs. Use when Lennart asks about cash flow, cash runway, when we run out of money, upcoming payments, receivables, or wants to see the forward cash position.
Triggers: "Cash flow forecast", "cash runway"
Data sources: Qonto, Moco, Close
Builds a rolling 13-week (3-month) cash flow forecast by combining actual bank data with known obligations and expected revenue. This is the forward-looking companion to the top-management-report's cash position snapshot.
Use this skill when Lennart says:
Always read before analysis:
context/finance.md -- Cost structure, billing rates, margin targets, financial calendarcontext/financial-planning.md -- Cash flow rules (green/yellow/red thresholds), revenue replacement model, cash flow leverscontext/current-priorities.md -- Strategic context (Neudorff cliff, pipeline priorities)https://thirdparty.qonto.com/v2Authorization: {slug}:{key}GET /organization -- account balancesGET /transactions?iban={iban}&settled_at_from=YYYY-MM-DD&settled_at_to=YYYY-MM-DD&per_page=100&sort_by=settled_at:desc -- transaction historyhttps://h2wconsultinggmbh.mocoapp.com/api/v1Authorization: Token token={MOCO_API_KEY}Report endpoints (use FIRST for pre-aggregated data):
GET /report/cashflow?from=&to= -- pre-aggregated cashflow transactions in account currencyGET /report/finance?from=&to= -- financial records (net amounts, types, by company)GET /account/fixed_costs?year=YYYY -- monthly fixed costs by category (payroll, rent, etc.)Invoice endpoints:
GET /invoices?status=created -- outstanding invoices (sent but not paid)GET /invoices?status=partially_paid -- partially paid invoicesGET /invoices?status=overdue -- overdue invoices...
CFO-level financial analysis with weekly dashboard, scenario modeling, deal economics, and investment decisions. Use when Lennart asks about margins, profitability, pricing, deal economics, revenue scenarios, cost analysis, investment decisions, financial health, CFO dashboard, weekly financial review, or wants a CFO perspective on any business decision.
Triggers: "Deal economics", "revenue scenarios", "CFO dashboard"
Data sources: Qonto, Close, Moco
Acts as a virtual CFO for Lennart. Goes beyond the top-management-report (which shows current position) to provide forward-looking financial analysis, scenario modeling, and decision support.
Use this skill when Lennart says:
Always read before analysis:
context/finance.md -- Entity economics, cost structure, billing rates, margin targets, unit economics. Includes actual rate realization data (EUR 148/hr blended, 87% at 145-155) and internal cost ratio trend.context/financial-planning.md -- Revenue replacement model, Labs build-up, cash flow rules, pricing rulescontext/current-priorities.md -- What's strategically important right nowcontext/goals.md -- Quarterly targetsdata/octaved-flow-export/ANALYSIS.md -- Historical baselines (Oct 2022 - Feb 2026): actual rates by customer, project profitability, utilization by person. Use for benchmarking deal economics against real data, not assumptions.context/clients.md -- Customer lifetime value table with historical hours, avg rates, and engagement duration per customerProduces a standalone PDF summarizing the financial state of the business. Designed to run weekly as part of the C-suite dashboard sequence.
Trigger: "CFO dashboard", "CFO weekly", or invoked by c-suite-dashboard orchestrator.
Data sources:
GET /organizationGET /transactions?iban={iban}&settled_at_from={first_of_month}GET /report/cashflow?from={date}&to={date} -- pre-aggregated cashflow dataGET /report/finance?from={date}&to={date} -- financial records (net amounts, types)...
"Check Claude usage limits via Playwright and cache the result. Use at session start, before heavy work, or when approaching limits. Also triggers auto-switch to OpenRouter if limits are critical."
Triggers: n/a
Data sources: n/a
Fetches real-time Claude usage data from claude.ai/settings/usage via Playwright, caches it locally, and triggers auto-switching if limits are critical.
Use when:
Use Playwright to navigate to https://claude.ai/settings/usage.
If redirected to login, tell Lennart: "Please log in to claude.ai in the Playwright browser window, then ask me to check usage again."
Use Playwright evaluate to call the API directly (faster than scraping):
async () => {
const resp = await fetch('/api/organizations/2efe1133-5a59-4f45-87c3-a1efbbe58382/usage');
return await resp.json();
}Take the JSON response and run:
cd "g:/My Drive/Lennart/Claude Projects/1-LHE Executive Assisant (EA)" && python scripts/model_manager.py cache-usage '<THE_JSON>'Replace <THE_JSON> with the actual JSON string from Step 2. Use single quotes around the JSON.
Run:
cd "g:/My Drive/Lennart/Claude Projects/1-LHE Executive Assisant (EA)" && python scripts/model_manager.py autoSummarize concisely:
Example output:
Claude usage: Session 85% (resets in 2h), Weekly 12%. Currently on
claudeprofile.OpenRouter: $9.73 remaining. Approaching session limit — consider switching before 100%.
Or if auto-switched:
Claude usage: Session 100% (resets in 1h 56m), Weekly 6%.
Auto-switched to
paid(OpenRouter). Balance: $9.73. Restart the session to apply.
Or if balance is low:
Claude usage: Session 100%, Weekly 6%. OpenRouter balance: $3.12 — LOW!
Latency will be degraded. Top up at openrouter.ai. Switched to
free(zero cost).
CHRO-level people strategy with weekly dashboard, retention risk assessment, succession planning, comp benchmarking, and org design. Use when Lennart asks about team health, retention risks, bench cost, redeployment, key-person dependencies, hiring decisions, org structure, comp analysis, or wants a people leadership perspective.
Triggers: "Team health", "retention risks", "CHRO dashboard"
Data sources: Moco, Jira, Hedy, Apollo, data/people/
Acts as a virtual Chief People Officer for Lennart. Owns team health, retention, capacity planning, development, and organizational design. Complements workforce-planning (which focuses on staffing scenarios) and evaluation-cycle (which is per-person) with a strategic people portfolio view.
Use this skill when Lennart says:
Always read before analysis:
context/team.md -- Team structure, key peoplecontext/team-assignments.md -- Who is on which projects (from Moco)context/team-capabilities.md -- Skills matrix, capacity profile, key-person dependenciescontext/team-history.md -- Join/departure dates, communication patternscontext/finance.md -- Salary cost structure (EUR 85-95K/month for 20 team members)context/risks.md -- Key-person dependencies (Risk #3), single-threaded accounts (Risk #11)context/current-priorities.md -- Neudorff redeployment contextdata/people/ -- Individual people records (evaluation history, incidents, development goals)plans/neudorff-redeployment.md -- Redeployment plan for exposed consultantsProduces a standalone PDF summarizing people and team health. Designed to run weekly as part of the C-suite dashboard sequence.
Trigger: "CHRO dashboard", "CHRO weekly", or invoked by c-suite-dashboard orchestrator.
Data sources:
GET /report/utilization?from=&to= -- daily billable/unbillable per user WITH target hoursGET /report/absences?year=YYYY&active=true -- per-user vacation (total/used/planned) + sick daysGET /users/{id}/performance_report?year=YYYY -- annual+monthly: target, tracked, billable, variation per userGET /users (headcount), GET /planning_entries?period=from:to (planned hours)...
Create, update, or deploy password-protected client microsites. Use when Lennart says create a client website, build a site for [company], positioning website, proposal site, workshop results site, QBR site, update the [client] website, deploy the [client] site, or client microsite.
Triggers: n/a
Data sources: n/a
Create branded, password-protected HTML microsites for prospects and customers. Single-file output, self-contained, deployed to {client-slug}.blconsultingservice.com.
| Mode | Trigger | Action |
|---|---|---|
| create | "create a client website for [company]" | Full workflow: brand scrape, content, generate, deploy |
| update | "update the [client] website" | Modify existing generator, regenerate, redeploy |
| deploy | "deploy the [client] site" | Run generator + push to Netlify |
| Type | When | Sections |
|---|---|---|
| positioning | Late-stage prospect evaluation | Summary, Analysis, Solutions, Zielbild, Roadmap, Business Case, Why H2W, References, Options, Next Step + Offer CTA |
| proposal | Alongside or replacing PDF proposal | Scope, Approach, Team, References, Commercial, Offer CTA |
| workshop | Post-workshop deliverable | Context, Findings, Recommendations, Quick Wins, Next Steps |
| qbr | Quarterly account review for existing customer | Delivery Summary, KPIs, Budget Status, Expansion, Action Items |
The canonical pattern is the MeisterWerke positioning site: reports/client-facing/positioning/meisterwerke-schulte/website/generate_site.py
Every new site follows this architecture. Read it before generating any new site.
1. Ask for company name (or infer from conversation context)
2. Look up in Close CRM (lead_search) and Moco (companies API)
3. Determine:
- Entity: H2W Consulting GmbH (DACH) / BLCG International (US) / H2W Labs
- Language: de (DACH/EU) / en (US/UK/international)
- Client slug: lowercase-hyphenated (e.g., "ott-hydromet")
4. Check if site already exists: reports/client-facing/{type}/{client-slug}/website/Ask Lennart which type: positioning / proposal / workshop / qbr. Default to positioning for prospects, qbr for existing customers.
1. Check brand/customer-logos/{client-slug}.png -- use if exists
2. If no logo: scrape client website homepage for logo URL, download to brand/customer-logos/
3. Scrape client website for:
- Primary brand color (for --c-accent CSS variable)
- Hero-quality lifestyle/product image (for hero background)
..."Autonomous end-of-day closeout: git commit, Moco time logging, CRM sync, daily log, session briefing, self-validation. Use when Lennart says closeout, wrap up, end session, done for now, save state. Replaces session-closeout."
Triggers: "Wrap up", "end session", "closeout", "done for now", "save state". Modes: --quick (git+briefing+time-split), --dry-run (preview). Replaces session-closeout.
Data sources: M365, Moco, Close, git, data/session-briefing.md, data/daily-log/, data/closeout-audit/
Single-command end-of-day closeout. Runs 8 steps autonomously without prompting. Every step produces a confirmation line. No step may be silently skipped.
| Flag | Steps | API writes | Asks questions |
|---|---|---|---|
| (none) | 1-8 | Moco, Close, git | No (fully autonomous) |
--quick | 1-3 | git only | No |
--dry-run | 1-8 | None (shows what would happen) | No |
Parse mode from Lennart's message. Examples: "closeout --dry-run", "quick closeout", "wrap up dry run".
Each step runs independently. A failure in one step does NOT abort the sequence. Failures are logged to the audit file and surfaced in Step 8 validation. In --dry-run mode, all API calls become "WOULD POST: [payload]" entries.
git status (never use -uall).git diff to see staged and unstaged changes.a. Stage relevant files by name (never git add -A; exclude .env, credentials, .obsidian/workspace.json). b. Draft a descriptive commit message from session activity. c. Commit. If pre-commit hook fails, fix and create a NEW commit (never amend).
Output: [1] Git: {N files committed (hash) | clean tree}
data/session-briefing.md and update:data/issues/ IDs)Output: [2] Session briefing: updated, {N} open loops
...
CMO-level marketing analysis with weekly dashboard, content ROI, channel effectiveness, and brand health. Use when Lennart asks about marketing performance, content output, lead attribution, brand visibility, content ROI, channel effectiveness, or wants a marketing leadership perspective.
Triggers: "Marketing performance", "content output", "CMO dashboard"
Data sources: marketing.md, Close, M365, Web
Acts as a virtual Chief Marketing Officer for Lennart. Owns content performance, lead attribution, channel effectiveness, and brand health. Complements marketing-content-pipeline (which creates content) with a strategic marketing performance view.
Use this skill when Lennart says:
Always read before analysis:
context/marketing.md -- Content pillars, channels, inventory, campaigns, metrics targets, inbound marketing system overviewcontext/current-priorities.md -- Strategic context (Neudorff case study deadline)context/goals.md -- Quarterly targetscontext/pipeline.md -- Pipeline for lead source attributiondata/content-queue/queue.md -- Content production queue state and buffer healthdata/content-queue/published/ -- Published content with performance datacontext/seo-keywords.md -- SEO keyword targets and ranking trackingProduces a standalone PDF summarizing marketing performance. Designed to run weekly as part of the C-suite dashboard sequence. Has a practical core (measurable today) and an aspirational roadmap (what to measure next).
Trigger: "CMO dashboard", "CMO weekly", or invoked by c-suite-dashboard orchestrator.
Data sources:
lead_search for new leads this week (check for source tags), find_opportunities for inbound-attributed dealsreports/snapshots/ for trend dataExecution:
...
Competitive landscape mapping, deal battle cards, win/loss pattern analysis, and comparison content generation. Use when Lennart says competitor analysis, battle card, competitive matrix, win/loss analysis, comparison page, who are we up against, competitive landscape, or differentiation strategy.
Triggers: "Competitor analysis", "battle card", "win/loss analysis"
Data sources: Apollo, Close, Hedy, Web
Systematic competitive intelligence gathering, analysis, and activation. Tracks competitors across the Infor LN consulting and AI/manufacturing markets, feeds battle cards into sales execution, analyzes win/loss patterns from CRM, and produces comparison content for marketing.
Use this skill when Lennart says:
Always read before analysis:
context/competitors.md -- Current competitive landscape and intelligence gapscontext/offerings.md -- Our service and product catalog with pricingcontext/win-loss.md -- Deal outcome patterns, pricing intelligencecontext/icp.md -- Ideal Customer Profiles, buyer personascontext/strategy.md -- Growth thesis, market position, competitive moatcontext/sales-playbook.md -- Sales stages, objection handlingLoad on demand:
context/clients.md -- Customer profiles for reference case matchingcontext/pipeline.md -- Active pipeline for deal intelligence context| Source | What It Provides | Access |
|---|---|---|
| Apollo.io | Company enrichment, job postings (hiring = investment signal), org data, headcount | MCP: apollo_organizations_enrich, apollo_organizations_job_postings, apollo_mixed_people_api_search |
| Close CRM | Win/loss data, deal notes, competitor mentions in activities, deal outcomes | MCP: activity_search, find_opportunities, fetch_opportunity, lead_search |
| Web research | Competitor websites, press releases, case studies, Infor partner directory | WebSearch, WebFetch |
| Hedy transcripts | Competitor mentions in client meetings, client perceptions | MCP or REST API (references/api-reference.md) |
| Context files | Baseline competitive data, our positioning, win/loss patterns | Local files |
Delegate data gathering to subagents for cost efficiency. Main conversation handles analysis, synthesis, and output.
Parallelization pattern:
...
Content performance tracking, weekly and monthly analytics reports, and pipeline attribution. Use when Lennart says content analytics, content performance, what's working, weekly content report, or attribution report.
Triggers: "Content analytics", "weekly content report"
Data sources: Published content, Close CRM
Tracks content performance across LinkedIn, blog, and newsletter. Produces weekly and monthly reports. Maps content to pipeline attribution when possible.
Use this skill when Lennart says:
Also proactively suggest:
Always read:
data/content-queue/published/ -- Published items with performance datadata/content-queue/queue.md -- Production metrics (items drafted, approved, published)context/marketing.md -- Targets and benchmarkscontext/seo-keywords.md -- Keyword ranking trackingTrigger: "Weekly content report", "How did content perform this week?"
Data collection:
data/content-queue/published/Output:
WEEKLY CONTENT REPORT -- Week of [Date]
PRODUCTION
- LinkedIn posts published: [X] / 21 target ([X]%)
- Blog articles published: [X] / 7 target ([X]%)
- Buffer health: LinkedIn [X] approved, Blog [X] approved
PERFORMANCE (if data available)
| Post | Date | Type | Language | Impressions | Engagement | Pillar |
|------|------|------|----------|-------------|------------|--------|
| [title] | [date] | LinkedIn | EN | [X] | [X]% | ERP |
Top performer: [title] -- [why it worked]
Underperformer: [title] -- [hypothesis for low performance]
CONTENT MIX
- By pillar: ERP [X], AI [X], Operations [X], Culture [X]
- By language: EN [X], DE [X]
- By slot: Morning [X], Midday [X], Afternoon [X]
VOICE CALIBRATION (weeks 1-4)
- Posts approved as-is: [X]
- Posts edited: [X]
- Posts rejected: [X]
- Key patterns: [summary from calibration log]
RECOMMENDATIONS
- [1-2 specific actions based on this week's data]...
Content production orchestrator. Checks queue health, drafts LinkedIn posts and blog articles in batch, manages the 3-day approved buffer, and stages content for review. Use when Lennart says content engine, fill the queue, draft posts, batch content, what needs drafting, or queue status.
Triggers: "Fill the queue", "draft posts", "queue status"
Data sources: Content queue, context files
Daily content production orchestrator. Checks the queue, identifies what needs drafting, produces batches of LinkedIn posts and blog articles, and stages everything for review via content-review.
This is the production workhorse. Where marketing-content-pipeline handles individual content requests, content-engine runs the system: checking buffer health, maintaining posting rhythm, and ensuring Lennart always has content ready to publish.
Use this skill when Lennart says:
Also proactively suggest this skill:
Always read before execution:
data/content-queue/queue.md -- Current queue state and buffer statuscontext/marketing.md -- Content pillars, backlog ideascontext/seo-keywords.md -- SEO targets for blog articlescontext/current-priorities.md -- What's top of mind (content should reflect this)context/offerings.md -- Product/service details for accuracybrand/brand-guide.md -- Brand colors, entity rules, image generation guidelinesQuick status of the content production system.
Trigger: "Queue status", "How's the buffer?", "Content status"
Execution:
data/content-queue/queue.mdOutput:
CONTENT QUEUE STATUS -- [Date]
Buffer Health:
- LinkedIn: [X] approved posts ready ([GREEN/YELLOW/RED], target: 9)
- Blog: [X] approved articles ready ([GREEN/YELLOW/RED], target: 3)
Queue Summary:
- Raw inputs waiting: [X]
- Drafts awaiting review: [X]
- Approved and ready: [X]
- Published this week: [X]
[Alert if buffer is YELLOW or RED with recommended action]...
Batch content review workflow. Presents drafted LinkedIn posts and blog articles for Lennart's review, tracks edits for voice calibration, and moves approved content to the publishing queue. Use when Lennart says review content, check drafts, approve posts, content review, or what's ready to review.
Triggers: "Review content", "approve posts", "content review"
Data sources: Content queue drafts
Presents batched content drafts for Lennart's review. Tracks his edits and rejections to calibrate voice over time. Moves approved content to the publishing queue and rejected content back with feedback.
This is Lennart's daily touchpoint with the content system. Designed for his ~30-minute morning review session.
Use this skill when Lennart says:
Also proactively suggest this skill:
Always read before execution:
data/content-queue/queue.md -- Queue statedata/content-queue/drafts/linkedin/ -- LinkedIn drafts awaiting reviewdata/content-queue/drafts/blog/ -- Blog drafts awaiting reviewdata/content-queue/drafts/linkedin/ and data/content-queue/drafts/blog/Present all LinkedIn drafts in a compact format:
LINKEDIN DRAFTS FOR REVIEW -- [Date]
[X] posts ready for review. Target dates: [dates]
---
POST 1 of [X] | [Morning/Midday/Afternoon] | [EN/DE] | Target: [date]
Pillar: [pillar] | ICP: [icp]
[Full post text]
---
POST 2 of [X] | ...After presenting all posts, ask:
If blog drafts exist, present after LinkedIn:
BLOG DRAFT FOR REVIEW -- [Date]
Title: [title]
Language: [EN/DE] | Type: [article type] | Target keyword: [keyword]
Word count: [count] | Meta description: [meta]
[Full article text]
---
Approve / Edit / Reject?For each reviewed item:
Approved:
drafts/[type]/ to approved/status: approved, add approved_date: YYYY-MM-DDEdited:
...
Refresh context files with live data from Moco, Close CRM, and Qonto. Run weekly (before Monday reports) or on demand to keep context files current. Use when Lennart says refresh context, update context, sync data, pull latest numbers, or before generating reports that depend on current data.
Triggers: "Refresh context", "pull latest numbers"
Data sources: Moco, Close, Qonto
Pulls live data from connected systems and updates the context files that all other skills depend on. Without regular refreshes, skills operate on stale data.
Use when Lennart says:
Updates all context files from all sources. Takes 2-3 minutes.
Updates pipeline.md and checks cash position. Use before financial reports.
Updates projects.md and team-assignments.md from Moco. Use before staffing discussions.
Which files to refresh is driven by context/index.md. Every entry has YAML frontmatter with auto_refresh and (when true) source and refresh_interval_days.
Step 0 (always): Read context/index.md. Build the refresh list: every entry where auto_refresh: true AND file age > refresh_interval_days. Files with auto_refresh: false are manually curated — NEVER overwrite them.
Current auto-refreshable files (for reference, but always re-derive from index.md so new files are picked up automatically):
| Source (frontmatter) | Target File | Handler |
|---|---|---|
moco | context/projects.md | Step 2 — Moco /projects |
moco | context/team-assignments.md | Step 3 — Moco /users + project assignments |
close-crm+moco | context/pipeline.md | Step 4 — Close CRM find_opportunities |
manual | context/clients.md, context/contracts.md, context/risks.md | Flag as "manual update needed" — do not auto-write |
Additional Moco endpoints consumed live by downstream skills (not written to context files): /report/utilization, /report/absences, /account/hourly_rates, /account/internal_hourly_rates. Qonto cash position feeds the weekly snapshot, not context.
When adding a new auto-refreshed context file: add the entry to context/index.md with auto_refresh: true + source + refresh_interval_days, then add a handler step below if the source is new.
For each entry in context/index.md with auto_refresh: true:
...
[triggers: improve copy, review this draft, make this more compelling, copy review, sharpen the messaging, proposal copy, landing page copy]
Triggers: "Improve copy", "copy review", "sharpen messaging"
Data sources: Context files
Improve the quality, clarity, and persuasiveness of all outbound copy. Applies to proposals, outreach emails, one-pagers, LinkedIn posts, landing pages, and any text that needs to move a manufacturing executive toward a decision.
Use this skill when Lennart says:
Before writing any new copy, build a brief. This is a hard gate: the brief must be confirmed by Lennart before writing begins.
Copy Brief Template:
| Field | Description |
|---|---|
| Purpose | What this piece needs to accomplish (e.g., get a discovery call booked, move from proposal to signed SOW) |
| Audience | Specific persona and seniority (e.g., CIO at a mid-market packaging manufacturer) |
| Core Problem | The operational pain this audience feels in their own words |
| Differentiation | Why us, not a competitor or doing nothing. What makes this engagement different. |
| Proof Available | Case studies, metrics, named references, certifications we can cite. No fabrication. |
| CTA | The single next step we want the reader to take |
| Context | Related materials: meeting notes, existing proposals, prior emails, one-pagers in brand/one-pagers/ |
| Language | German (Sie/Du) or English. For German, check Du/Sie per email-drafting rules. |
| Entity | BLCG International, H2W Consulting GmbH, BL Consulting UG, or H2W Labs |
If Lennart provides enough context to fill the brief implicitly, fill it and confirm. Do not ask questions that the context already answers.
Apply the Feature-Benefit-Outcome framework to existing copy. For every claim or statement in the draft:
Review checklist:
...
[triggers: edit this, review this copy, proofread, quality check, polish this draft, improve this text, copy edit, seven sweep]
Triggers: "Edit this", "proofread", "quality check", "seven sweep"
Data sources: Context files
Structured quality editing pass for all written output: proposals, outreach emails, blog posts, LinkedIn content, client reports.
Read before every edit:
.claude/rules/communication-style.md (formatting, tone, no emojis, no em dashes).claude/rules/sales-and-proposals.md (proposal structure, value-led writing).claude/rules/email-drafting.md (Du/Sie, pre-send checklist, no filler)Apply all 7 sweeps sequentially. Use for high-stakes content: proposals, case studies, important outreach, client-facing reports.
Apply Sweep 1 (Clarity) and Sweep 3 (So What) only. Use for routine content: LinkedIn posts, internal emails, blog drafts, status updates.
Verify content matches the target tone. Cross-reference against communication-style.md rules.
Checks:
...
Run CRM and Moco data quality validation. Use when Lennart says validate CRM, check data quality, run CRM audit, or after any Close/Moco write operation.
Triggers: n/a
Data sources: n/a
Validates CRM and Moco data quality, auto-corrects where possible, alerts for manual intervention.
Use when:
python scripts/crm_validators/runner.py --dry-runReview output, then run without --dry-run to apply heals.
python scripts/crm_validators/runner.py --validator pipeline_stagefrom scripts.crm_validators import validate_pipeline_stage, validate_contact_enrichment
result = validate_pipeline_stage(opp_data)
if result.needs_attention:
# Present to LennartResults saved to data/crm-validation/YYYY-MM-DD.json (or -dry.json for dry runs).
CRO-level revenue analysis with weekly dashboard, deal intelligence, revenue forecasting, and sales performance. Use when Lennart asks about pipeline health, revenue forecast, sales velocity, deal prioritization, conversion rates, CRO dashboard, or wants a revenue leadership perspective on pipeline and sales execution.
Triggers: "Pipeline health", "revenue forecast", "CRO dashboard"
Data sources: Close, Moco, M365, snapshots
Acts as a virtual Chief Revenue Officer for Lennart. Owns pipeline health, sales execution, conversion analytics, and revenue forecasting. Complements sales-intelligence (which is deal-level tactical) with a strategic revenue leadership view.
Use this skill when Lennart says:
Always read before analysis:
context/pipeline.md -- Current pipeline snapshotcontext/finance.md -- Revenue baseline, rate realizationcontext/financial-planning.md -- Revenue replacement model, Neudorff scenarioscontext/current-priorities.md -- Strategic contextcontext/goals.md -- Quarterly targetscontext/sales-playbook.md -- Sales stages, discovery processcontext/deal-patterns.md -- Historical win/loss patterns, sales cycle benchmarkscontext/win-loss.md -- Deal outcome patternsdata/octaved-flow-export/ANALYSIS.md -- Historical revenue by customer for baselineProduces a standalone PDF summarizing revenue and pipeline health. Designed to run weekly as part of the C-suite dashboard sequence.
Trigger: "CRO dashboard", "CRO weekly", or invoked by c-suite-dashboard orchestrator.
Data sources:
find_opportunities (active, won, lost), find_pipelines_and_opportunity_statuses, activity_search, aggregation, org_usersbash scripts/m365.sh sent 7 for outbound activity matchingGET /projects?include_archived=false for current revenue baselinereports/snapshots/ for trend dataExecution:
...
CSO-level strategic analysis with weekly dashboard, competitive intelligence, strategic decision support, and market positioning review. Use when Lennart asks about strategic bets, risk register, competitive moves, market positioning, strategic decisions, quarterly goal tracking, or wants a strategy leadership perspective.
Triggers: "Strategic bets", "competitive intel", "CSO dashboard"
Data sources: Context files, Close, Apollo, Web
Acts as a virtual Chief Strategy Officer for Lennart. Owns strategic monitoring, competitive intelligence, initiative tracking, and long-term positioning. Complements strategic-review (which is quarterly deep-dive) with a weekly pulse on strategic health.
Use this skill when Lennart says:
Always read before analysis:
context/strategy.md -- Growth thesis, market position, strategic bets, competitive moatcontext/risks.md -- Full risk register with statuscontext/goals.md -- Quarterly targetscontext/current-priorities.md -- Current focus areascontext/competitors.md -- Competitive landscapecontext/icp.md -- Ideal Customer Profilescontext/offerings.md -- Service and product catalogcontext/clients.md -- Customer portfolio for concentration analysisProduces a standalone PDF summarizing strategic health. Designed to run weekly as part of the C-suite dashboard sequence.
Trigger: "CSO dashboard", "CSO weekly", or invoked by c-suite-dashboard orchestrator.
Data sources:
find_opportunities for deal type distribution (ERP vs. Labs, new vs. existing)reports/snapshots/ for concentration trendsExecution:
PDF sections:
...
Build deep, comprehensive customer profiles by pulling from all systems (Close CRM, Jira, Moco, Hedy, Apollo.io, web). Use when Lennart asks to understand a customer deeply, build a customer profile, analyze customer health, map stakeholders, find expansion opportunities, or review the full customer portfolio.
Triggers: "Customer profile", "stakeholder map"
Data sources: Close, Jira, Moco, Hedy, Apollo
Build and maintain deep customer understanding by pulling data from every available system, synthesizing it into a comprehensive profile, and identifying strategic opportunities and risks.
This goes beyond meeting prep or deal status. It builds a living, 360-degree view of each customer relationship.
Use this skill when Lennart says:
lead_search / search / fetch_lead -- full lead record, contacts, custom fieldsfind_opportunities / fetch_opportunity -- all deals (open, won, lost)activity_search -- full activity history (emails, calls, meetings, notes)aggregation -- activity volume, deal metricssearchJiraIssuesUsingJql -- all issues by customer (cloudId: 0ee5c601-cc7e-4022-ba05-749fee885399)getJiraIssue -- issue detailssearchConfluenceUsingCql / getConfluencePage -- project documentationAuthorization: Token token=$MOCO_API_KEYhttps://h2wconsultinggmbh.mocoapp.com/api/v1Report endpoints (use FIRST, pre-calculated):
GET /projects/{id}/report -- budget, hours, costs, profitability per project (single call, replaces manual aggregation)GET /account/hourly_rates?company_id={id} -- customer-specific billing ratesRaw data endpoints:
GET /projects -- project budgets, timelines, statusGET /projects/{id} -- project details with assigned staffGET /activities -- time entries and effort data (use when project-level breakdown needed)GET /invoices -- invoicing historyGetSessions -- find meetings involving the customerGetSessionDetails -- meeting summaries, action items, decisionsapollo_organizations_enrich -- company data, technographicsapollo_mixed_people_api_search -- find key personas...
Cross-functional daily business scan. Use when Lennart says what is on my calendar, anything new, what should I know, daily scan, or at session start. Covers all 6 business areas (sales, delivery, people, finance, operations, product).
Triggers: "What's on my calendar", "daily scan", "calendar check"
Data sources: M365, Close, Hedy
Cross-functional scan across all business areas. Pulls live data from M365, Close CRM, Moco, and Jira, then presents a unified "what needs attention" view organized by business function.
Use this skill when Lennart says:
Pull live data from all sources BEFORE reading local files.
bash scripts/m365.sh inbox-scan 15 # H2W account
bash scripts/m365.sh inbox-scan 15 --account blc # BLC accountFrom results:
"importance": "high" emails first with [!] markermail-read <id>find_opportunities(status_type="active", updated_at={"start": "<24h ago>"}, sort_by="largest_value")
find_opportunities(status_type="won", close_at={"start": "<7 days ago>"})Flag: stage changes, confidence changes, Won/Lost since last session.
# Recent offers
curl -s -H "Authorization: Token token=$MOCO_API_KEY" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/offers?sort=-date&per_page=10"
# Active client projects with budget data (exclude internal)
curl -s -H "Authorization: Token token=$MOCO_API_KEY" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/projects?active=true&per_page=100" \
| python -c "import json,sys; [print(json.dumps({k:p[k] for k in ['id','name','customer','budget','budget_expenses','billable']})) for p in json.load(sys.stdin) if p.get('billable')]"
# Projects updated since yesterday
curl -s -H "Authorization: Token token=$MOCO_API_KEY" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/projects?updated_after=<yesterday>T00:00:00Z&sort=-updated_at&per_page=10"
..."Create a daily log entry summarizing EA project activity and company business events. Use when Lennart says daily log, log today, what happened today, end of day, or automatically via evening cron."
Triggers: n/a
Data sources: n/a
Capture what happened today across both the EA project (tooling, skills, infrastructure) and the business (sales, finance, operations, delivery, team).
Pull from all available sources to build a complete picture:
decisions/log.md) -- filter today's entriesdata/session-briefing.md) -- last session summary and open loopsdata/draft-log.md) -- emails drafted todaygit log --oneline --since="YYYY-MM-DDT00:00:00" --until="YYYY-MM-DDT23:59:59"bash scripts/m365.sh inbox-scan 20 for notable inbound emailsbash scripts/m365.sh calendar for meetings that happened todaybash scripts/m365.sh sent --after YYYY-MM-DD --before YYYY-MM-DD+1 for emails actually sentdata/issues/) -- any issues created, updated, or closed todaydata/content-queue/) -- content drafted, approved, or published todayWrite to data/daily-log/YYYY-MM-DD.md:
# Daily Log: [Day of Week], [Month DD, YYYY]
## Business Activity
### Sales
- [Deal movements, proposals sent, meetings held, pipeline changes]
### Finance
- [Cash position updates, invoices, payments, intercompany transfers]
### Delivery
- [Project updates, escalations, client meetings, milestones]
### Team / Operations
- [Directives, HR items, process changes, internal meetings]
### Meetings
- [Calendar events that occurred today with key outcomes if known]
## EA Project Activity
- [Skills built/modified, infrastructure changes, tooling fixes, context file updates]
## Emails
### Sent
- [Emails actually sent today]
### Drafted (pending review)
- [Drafts created but not yet sent]
### Notable Inbound
- [Important emails received that need attention]
## Content
- [LinkedIn posts, blog articles drafted/approved/published]
## Decisions Logged
- [Count and brief list from decisions/log.md]
## Open Items (End of Day)
...[triggers: dashboard layout, KPI design, improve this report, report design, dashboard design, better visualization, report layout]
Triggers: "Dashboard layout", "KPI design", "report design"
Data sources: report-design-standards
Design, review, and optimize PDF dashboard reports generated via reports/report_engine.py.
Design the layout for a new PDF dashboard. Select KPIs, arrange sections, choose visualization types.
Process:
context/kpis.mdreport_engine.py implementationOutput: Section-by-section layout spec with method calls, data sources, and column definitions.
For a given audience and purpose, recommend which KPIs to display.
Process:
context/kpis.md for the full KPI catalog across 6 categories (Financial, Utilization, Sales, Delivery, Customer, Labs)Output: KPI specification table with data sources and RAG thresholds.
Review an existing PDF report for design quality and recommend improvements.
Checklist:
references/report-design-standards.md...
Score all qualified pipeline opportunities (Interested but need time through Verge of closing) on a 0-100 health scale and tier them Green/Yellow/Red/Stalled. Use when Lennart asks to score deals, check deal health, find stuck or at-risk opportunities, or run the weekly pipeline health review. Writes results to Close custom fields for Smart View filtering.
Triggers: n/a
Data sources: n/a
Rule-based scoring for the qualified sales funnel. Signal-driven, stage-aware, idempotent.
Scores all active opportunities in these stages (~90 opps across Infor LN + Development pipelines):
Explicitly excluded: Diverse conversation, Potential for Discussion, and all Closed Won/Lost. Those belong to lead scoring (separate skill).
Score starts at 100. Penalties subtract for missing/stale signals. See scripts/deal_health_score_config.py for the full weights table. Signals:
Tiers:
python scripts/deal_health_score.py --dry-runPrints sorted table to stdout, writes JSON summary to data/deal-health/YYYY-MM-DD-dry.json. Use this to preview before any CRM updates.
python scripts/deal_health_score.py --opp oppo_XXXXXXXXpython scripts/deal_health_score.pyScores all in-scope opps and writes 4 custom fields back to Close:
Also writes data/deal-health/YYYY-MM-DD.json for downstream skills.
python scripts/deal_health_score.py --dry-run --limit 10cf_4nvJ6NDyMzQMG4Z94lghkQuQgbxiPAusguMYlqdBX9Hcf_qsdqIOFyUyeWLFtBhhgv8FHir1ejkGMvjZhaCjuSTly...
Sales execution loop that monitors email and calendar for deal signals, flags overdue follow-ups, and surfaces deals needing action. Use when Lennart says deal pulse, check my deals, any replies, follow-up check, what needs attention in the pipeline, or sales check-in.
Triggers: "Deal pulse", "check my deals", "follow-up check"
Data sources: M365, Close
Monitors email, calendar, and CRM to surface deals that need attention right now. Combines email signal detection, overdue follow-up tracking, and calendar-to-pipeline cross-referencing into one actionable view.
This is the daily sales execution companion. Where sales-intelligence gives you the full pipeline picture, deal-pulse tells you "what do I need to do today to keep deals moving?"
Use this skill when Lennart says:
Also proactively suggest this skill:
Check data/deal-health/ for today's scoring JSON. If missing or >24h old, run:
python scripts/deal_health_score.pyUse the tier/score/reason from the JSON to prioritize. Red + Stalled deals get scanned first, Yellow second, Green only if time permits or user explicitly asks. The tier/reason replaces the old "overdue follow-up" heuristic.
Pull active opportunities from Close CRM:
find_opportunities with status_type "active"org_users to map user IDs to namesFor each active deal (or top 10 by value if many), search email:
bash scripts/m365.sh mail-search "[company name]" --account h2w
bash scripts/m365.sh mail-search "[company name]" --account blcMANDATORY: Always scan BOTH mailboxes (h2w AND blc) for every deal. Never default to h2w only. BLC-contracted deals (US clients, Trojan, AMS, OTT, Calanbau, Containex signature workflows via Docusign) route critical correspondence through the BLC account. A single-mailbox scan produces misleading follow-up drafts because you miss: contract/DocuSign status, warm-call summaries, real customer contacts, delivery bounces. Before drafting any follow-up, both searches must have been run and their results merged.
...
"Check for unsent email drafts and remind about stale ones. Use when Lennart says check drafts, pending drafts, draft status, unsent emails, what drafts are waiting, or did I send that email."
Triggers: "Check drafts", "pending drafts", "unsent emails"
Data sources: data/draft-log.md, M365 drafts
Check local draft log against Outlook drafts folder. Surface stale drafts with urgency tiers. Auto-clean sent/deleted drafts.
Use when Lennart says:
| Urgency | YELLOW | RED |
|---|---|---|
| prospect | 12 hours | 24 hours |
| client | 12 hours | 24 hours |
| partner | 24 hours | 48 hours |
| internal | 24 hours | 48 hours |
| low | 48 hours | 72 hours |
Read data/draft-log.md. If the file does not exist or has no data rows (only header), output:
DRAFT TRACKER -- [Date]
No tracked drafts. The log is empty.And stop.
Parse each data row into: ID, Account, Recipient, Subject, Created, Source, Urgency.
For each unique account in the log (h2w, blc), run:
bash scripts/m365.sh draft-list 50
bash scripts/m365.sh draft-list 50 --account blc # only if blc rows existImportant: Do NOT run both accounts concurrently. Run h2w first, then blc.
Collect all Outlook draft IDs into a set.
For each row in the draft log:
data/draft-log.md.Update data/draft-log.md with only the remaining rows (keep the full header: title, description line, blank line, table header, separator).
For each remaining draft, calculate age from Created timestamp to now (UTC).
Assign status based on urgency tier and age:
DRAFT TRACKER -- [Date]
PENDING DRAFTS: [total] ([stale] stale)
| # | Recipient | Subject | Age | Source | Status |
|---|-----------|---------|-----|--------|--------|
| 1 | [email] | [subject] | [Xd Xh] | [skill] | RED |
| 2 | [email] | [subject] | [Xh] | [skill] | YELLOW |
| 3 | [email] | [subject] | [Xh] | [skill] | GREEN |Sort by status: RED first, then YELLOW, then GREEN.
"Stale" count = YELLOW + RED drafts.
...
CRM contact enrichment procedures, custom field IDs, and email drafting checklists. Auto-loads when creating/updating contacts, drafting emails, or doing CRM enrichment work.
Triggers: n/a
Data sources: n/a
When creating or updating a contact in Close CRM, always populate all available fields before saving.
Every contact create or update must include both standard fields and custom fields in a single operation. Never do one without the other.
Step 1: Standard fields (via MCP create_contact / update_contact or REST API):
| Field | Required | Notes |
|---|---|---|
| name | Always | Full name |
| Always | Primary email | |
| title | Always | Job title (not company name) |
| phones | If available | From email signature or Apollo |
| urls | If available | LinkedIn URL as {"type": "url", "url": "..."} |
Step 2: Custom fields (via REST API, MCP tools do NOT support these):
export $(grep CLOSE_API_KEY .env)
curl -s -X PUT -u "$CLOSE_API_KEY:" -H "Content-Type: application/json" \
-d '{"custom.cf_FIELD_ID": "value", "custom.cf_OTHER": "value"}' \
"https://api.close.com/api/v1/contact/{contact_id}/"Set ALL of these when data is available. Use get_fields to verify valid choice values if unsure.
| Field | ID | Type | Source |
|---|---|---|---|
| Contact Role | cf_Ua5RgjVEOxxqDW8GohGdfOwHpMrET3SqQTvNnF9oBRX | choices: Decision Maker, Gatekeeper, Point of Contact | Infer from title |
| Contact Country | cf_7ZWFQ7ud3Juw2LHyYyULZuxA7BKAZbkT1hqjGDJsM1h | text | Email domain, signature, lead |
| Contact City | cf_LNr9I0SIf8gPdTOKQ8SDAInvj5Z8PARcsvrJyUdsKfg | text | Signature, Moco, lead |
...
Design and build email sequences for Close CRM. Sales follow-up, nurture, re-engagement, workshop invitations, Labs product intros, and multi-touch campaigns. Use when Lennart says build a sequence, design email cadence, create a drip campaign, audit my sequences, or campaign builder.
Triggers: "Build a sequence", "design email cadence", "drip campaign"
Data sources: Close, context files
Designs, builds, and audits email sequences in Close CRM for sales follow-up, lead nurturing, re-engagement, and marketing campaigns. Produces bilingual (German + English) sequences with timing rules aligned to the sales playbook cadences.
This skill DESIGNS and BUILDS sequences. deal-pulse detects signals that trigger sequences. sales-outreach drafts individual messages. This skill creates the automated multi-touch cadences that run in Close CRM.
Use this skill when Lennart says:
Also proactively suggest:
Always read before execution:
context/sales-playbook.md -- Follow-up cadences, objection handling, deal archetypes, real email patternscontext/offerings.md -- Service and product catalog for CTAs and value propscontext/icp.md -- Ideal customer profiles, buyer personas, fit scoringcontext/marketing.md -- Content pillars, newsletter structure, campaign ideascontext/clients.md -- Customer tiers and relationship history for segmentationcontext/communication-preferences.md -- Per-person channel preferencesRead on demand:
context/win-loss.md -- What resonates and what fails in our messagingcontext/competitors.md -- Competitive positioning for differentiation emailsbrand/one-pagers/README.md -- Available collateral to attach or referenceEvery email in every sequence must follow these rules.
Example:
...
"Run the full employee evaluation cycle: prep, agenda, and post-meeting record update. Use when Lennart says evaluation for [name], run evaluation cycle, performance review for [name], or after completing an evaluation meeting."
Triggers: "Evaluation for [name]", "performance review for [name]"
Data sources: Moco, M365, Jira, Hedy, data/people/
Full employee evaluation lifecycle: preparation, agenda generation, and post-meeting record update.
Use this skill when Lennart says:
Runs people-prep skill for the person, then adds evaluation-specific context:
GET /report/utilization for pre-calculated data)GET /users/{moco_id}/performance_report?year=YYYY for target vs tracked vs billable with monthly breakdownOutput: Full evaluation prep briefing (as shown in people-prep, plus comparative context and suggested agenda).
Generate a structured meeting agenda based on the prep:
EVALUATION AGENDA: [Name]
Date: [Scheduled date]
Duration: 45-60 minutes
1. OPENING (5 min)
- How are you feeling about your role?
- What's going well? What's frustrating?
2. ACCOUNT REVIEW (15-20 min)
[For each account they manage:]
- [Account]: [1-2 specific questions based on prep data]
3. PERFORMANCE DISCUSSION (10 min)
[If incidents exist:]
- [Incident]: What happened? What did you learn? What changed?
[If no incidents:]
- General feedback on strengths and areas to develop
4. DEVELOPMENT (10 min)
- Career goals: Where do you want to be in 12 months?
- Skills: What do you want to learn?
- Support: What do you need from [manager] and from Lennart?
5. OPERATIONAL (5 min)
- Moco time tracking: Explain project-level logging expectation
- [Any other process topics]
6. OPEN FLOOR (5 min)
- Anything else you want to raise?
NOTES FOR LENNART:
- [Sensitive point to handle carefully]
- [Context to keep in mind but not say directly]...
Run the H2W Labs sales conversion playbook for a specific deal. Generates Opportunity Briefs, demo prep, proposals, and follow-up cadences based on deal stage. Use when Lennart asks to move a Labs deal forward, generate an Opportunity Brief, prep a demo, or run the playbook for a specific company.
Triggers: "Run the playbook for [company]"
Data sources: Close, context files
Orchestrates the 30-day stage-gate sales process for H2W Labs deals (KnowKit, Product Assistant, EasyShare, e-invoicing, Edge Apps).
Scope: H2W Labs deals only. ERP consulting deals through BLCG / H2W Consulting follow a separate process.
Use this skill when Lennart says:
Four stages targeting 30 days from first meeting to signed deal.
Goal: Send a personalized Opportunity Brief within 48 hours.
Execution:
0ee5c601-cc7e-4022-ba05-749fee885399)templates/h2w-labs-opportunity-brief.mdOutput: Completed Opportunity Brief ready to send.
If no response in 5 days: Draft a follow-up email using the sales-outreach skill (Scenario 5: Follow-Up After Discovery Call).
Two paths based on deal value:
Path A: Tailored Demo (deals under EUR 10K)
templates/h2w-labs-demo-prep.mdPath B: Scoping Workshop (deals EUR 10K+)
...
Auto-tag Hedy sessions to topics and update the local session index. Called by daily-calendar-scan at end of each run. No manual involvement. Use when sessions need tagging, index needs updating, or as part of daily/context-refresh workflows.
Triggers: Auto-tags sessions to topics, updates index. Called by daily-calendar-scan and context-refresh.
Data sources: Hedy MCP/REST, M365 calendar, Close CRM, data/hedy-*.json
Tags new Hedy sessions to the correct customer/internal topic and updates the local session index. Fully automated, no manual input.
Called automatically by:
data/hedy-topics-map.json -- maps customer names to Hedy topic IDs, email domains to customers, and session context IDsdata/hedy-sessions-index.json -- maps session IDs to customer, date, meeting type, attendeesdata/hedy-sessions-index.json to get last_scan timestampGetSessions with limit 20 (MCP in Claude.ai) or REST API (Claude Code): export $(grep HEDY_API_KEY .env)
curl -s -H "Authorization: Bearer $HEDY_API_KEY" \
"https://api.hedy.bot/sessions?limit=20"last_scan that are not already in the indexlast_scan and exitGet calendar events covering the time range of untagged sessions:
# Determine date range from untagged sessions (earliest to latest)
bash scripts/m365.sh calendar-range <start_date> <end_date>
bash scripts/m365.sh calendar-range <start_date> <end_date> --account blcIf calendar-range is not available, use calendar N where N covers the needed range:
bash scripts/m365.sh calendar 14
bash scripts/m365.sh calendar 14 --account blcFor each untagged Hedy session:
startTime to calendar event start times"calendar_matched": false in the indexFrom the matched calendar event:
h2w-consulting.com, blconsultingservice.coma. Look up domain in data/hedy-topics-map.json -> domain_to_customer b. If domain match found: use that customer name
...
"Scan inbox for emails needing replies, draft responses for all of them. Runs as part of daily-calendar-scan or standalone. Use when Lennart asks to check inbox, draft replies, process emails, clear inbox, what needs a reply, or inbox drafts."
Triggers: "Inbox drafts", "draft my replies", "what needs a reply"
Data sources: M365, Close, context
Scan both H2W and BLC inboxes for emails that need a reply. Draft a response for each one in Outlook drafts. Lennart reviews and sends or deletes.
Use this skill when Lennart says:
An email needs a reply if ALL of these are true:
| Sender | Language | Tone |
|---|---|---|
| External client / prospect (DACH) | German | Formal ("Sehr geehrte/r...", "Mit freundlichen Gruessen") |
| External client / prospect (international) | English | Professional but warm |
| Internal team (Heiko, Diego, Richard, Uli, Lucia, Carlos, etc.) | German or English (match their email) | Casual, direct |
| Partners / freelancers | Match their language | Professional but collegial |
| Personal contacts | Match their language | Casual |
If you lack context to write a good reply (technical question you can't answer, decision that needs Lennart's input, sensitive topic):
...
"Track intercompany billing between BLC UG and H2W GmbH. Reconcile invoices against bank transfers, flag missing payments, detect entity confusion. Use when Lennart asks about intercompany status, BLC to H2W transfers, intercompany reconciliation, entity billing, or when cash position needs investigation."
Triggers: "Intercompany status", "BLC to H2W transfers"
Data sources: Qonto, Moco
Track BLC-to-H2W intercompany invoices, flag missing transfers, reconcile against Qonto transactions, and catch entity confusion. Critical for cash flow visibility given H2W GmbH cash position.
Use this skill when Lennart says:
Trigger: "Intercompany status", "BLC to H2W transfers" Current intercompany balance: what BLC owes H2W, pending transfers, last transfer date, aging.
Trigger: "Intercompany reconciliation", "reconcile intercompany" Match Moco invoices against Qonto transactions for both accounts. Flag discrepancies.
Trigger: "Intercompany audit", "entity billing check" Full audit: entity confusion detection, missing intercompany entries, billing pattern analysis.
H2W GmbH (receiving account):
source .env && curl -s -H "Authorization: $QONTO_H2W_AUTH" \
"https://thirdparty.qonto.com/v2/transactions?iban=$QONTO_H2W_IBAN&settled_at_from=YYYY-MM-DD&settled_at_to=YYYY-MM-DD"BLC UG (sending account):
source .env && curl -s -H "Authorization: $QONTO_BLC_AUTH" \
"https://thirdparty.qonto.com/v2/transactions?iban=$QONTO_BLC_IBAN&settled_at_from=YYYY-MM-DD&settled_at_to=YYYY-MM-DD"Important: Qonto returns balances in EUR directly. Do NOT divide by 100.
source .env && curl -s -H "Authorization: Token token=$MOCO_API_KEY" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/invoices"context/finance.md -- entity structure, billing flowreferences/intercompany-billing-cheat-sheet.md -- rules and entity flowPhase 1: Read Context
references/intercompany-billing-cheat-sheet.md for intercompany rulescontext/finance.md for entity structure...
Invoice aging report from Moco and Qonto. Shows outstanding receivables by customer and age bucket, overdue invoices, collection risk, and payment patterns. Use when Lennart asks about outstanding invoices, overdue payments, receivables, collections, who owes us money, or payment status.
Triggers: "Outstanding invoices", "who owes us"
Data sources: Moco, Qonto
Tracks outstanding receivables by age and customer to protect cash position. Combines Moco invoice data (what was sent) with Qonto transaction data (what was paid) to identify collection gaps and payment patterns.
Use this skill when Lennart says:
Triggered by: "invoice aging", "outstanding invoices", "receivables", "AR report" Shows: All open invoices grouped by age bucket (current, 30, 60, 90+ days) with customer totals.
Triggered by: "have they paid", "did [customer] pay", "payment status for [customer]" Shows: One customer's invoice and payment history.
Triggered by: "what's overdue", "who owes us", "collection report" Shows: Overdue invoices ranked by amount, with suggested collection actions.
https://h2wconsultinggmbh.mocoapp.com/api/v1Authorization: Token token={MOCO_API_KEY}GET /invoices -- all invoices (paginate at 100)GET /invoices?status=created -- sent but not paidGET /invoices?status=partially_paid -- partially collectedGET /invoices?status=paid -- fully paid (for payment pattern analysis)GET /invoices?from=YYYY-MM-DD&to=YYYY-MM-DD -- invoices by date rangehttps://thirdparty.qonto.com/v2GET /transactions?iban={iban}&settled_at_from=YYYY-MM-DD&settled_at_to=YYYY-MM-DD -- incoming paymentsPhase 1: Invoice Data (main conversation, curl)
source .env && export MOCO_API_KEY && export MOCO_BASE_URL...
"Track issues, tasks, and topics with internal and external people and companies. Use when Lennart says add issue, track this, open issues, issue list, prep for meeting with X, issue dashboard, update issue, close issue, what am I tracking, or any variant of managing cross-functional tasks."
Triggers: "Add issue", "track this", "open issues", "prep for [company]", "issue dashboard"
Data sources: data/issues/
Lightweight issue tracking for cross-functional topics, operational requests, relationship items, and meeting follow-ups. Covers everything that does not belong in Jira (delivery) or Close CRM (deals).
Use this skill when Lennart says:
Parse the user's intent into one of these commands. The argument format shown is for reference -- users will speak naturally, not use exact syntax.
Create a new issue. Gather these fields interactively (ask if not provided):
| Field | Required | Default |
|---|---|---|
| title | yes | -- |
| company | yes | -- |
| contacts | no | [] |
| priority | no | medium |
| type | no | task |
| assignee | no | Lennart Hector |
| due_date | no | none |
| tags | no | [] |
| related | no | [] |
| description | yes | -- |
Execution:
data/issues/.counter to get the next ID numberISS-XXX (zero-padded to 3 digits)data/issues/ISS-XXX-slug.md with YAML frontmatter and body:---
id: ISS-XXX
title: [title]
status: open
priority: [priority]
type: [type]
assignee: [assignee]
company: [company]
contacts: [contacts list]
due_date: [YYYY-MM-DD or null]
created: [today YYYY-MM-DD]
updated: [today YYYY-MM-DD]
tags: [tags list]
related: [related list]
---
## Description
[description]
## Updates
- [today]: Created..counter and saveUpdate an existing issue. User may want to:
Execution:
data/issues/- [today]: [note]updated to today's date...
"Prepare and structure H2W Labs demo and PoC deliveries. Generates demo scripts, environment checklists, success criteria, and post-demo follow-up. Use when Lennart asks to prep a demo, build a demo script, PoC delivery plan, demo checklist, or prepare a Labs product demonstration for any client. Maps to planned SOP-014."
Triggers: "Prep demo for [company]", "PoC delivery plan"
Data sources: Close, Hedy, context
Structure the delivery of H2W Labs demos and proof-of-concept projects. This skill covers the delivery side -- h2w-labs-playbook covers the sales side.
Use this skill when Lennart says:
Trigger: "Prep demo for [company]", "demo script" Full demo preparation package: script, environment checklist, success criteria, risk mitigation.
Trigger: "PoC delivery plan", "PoC scope for [company]" Structured PoC delivery plan: scope, timeline, resources, acceptance criteria, handoff to production.
Trigger: "Demo follow-up for [company]", "debrief the demo" Process demo outcomes: what worked, what didn't, next steps, proposal trigger.
lead_search / fetch_lead -- company record, contactsfind_opportunities -- Labs deal stage, historyactivity_search -- recent communicationsGetSessions -- find discovery/scoping meetings with this clientGetSessionDetails -- get full transcript (transcript or cleaned_transcript). Extract use cases, pain points, and requirements from the raw conversation. Do not rely on Hedy's recap or meeting_minutes.context/offerings.md -- H2W Labs product catalogcontext/clients.md -- client profile if existingtemplates/h2w-labs-demo-prep.md -- demo prep templatereferences/h2w-labs-crm-stages.md -- pipeline stagestemplates/h2w-labs-demo-prep.md for structurecontext/offerings.md for product detailsreferences/h2w-labs-crm-stages.md for deal stageLaunch parallel data gathering:
close-crm-agent: Full lead record, Labs opportunity details, recent activity
hedy-agent: Pull full transcript of discovery meeting. Return raw transcript for analysis (do not summarize).
DEMO PREPARATION: [Company Name]
Product: [KnowKit / Product Assistant / Edge App / Custom]
Date: [Demo date]
Audience: [Names, titles from CRM]
...Build and execute go-to-market launch plans for H2W Labs products. Positioning, messaging, reference cases, launch sequences, channel coordination. Use when Lennart says launch strategy, go-to-market, GTM plan, product launch, reference case, or product positioning for a Labs product.
Triggers: "Launch strategy", "GTM plan", "product positioning"
Data sources: Close, context files
Structured go-to-market playbook for launching H2W Labs products into the market. Where h2w-labs-playbook handles converting individual deals (1:1 sales), this skill handles market-level launches: positioning a product, building reference cases, coordinating multi-channel campaigns, and driving pipeline at scale.
Scope: H2W Labs products only (KnowKit, Product Assistant, EasyShare, Factory Track, E-Invoice Integration, future Edge Apps). ERP consulting services use separate sales processes.
Use this skill when Lennart says:
Read before execution:
context/offerings.md -- Product catalog, features, pricingcontext/icp.md -- Ideal Customer Profiles, buyer personascontext/clients.md -- Existing customers (reference case candidates)context/marketing.md -- Content pillars, channels, inventorycontext/strategy.md -- Growth thesis, market positioncontext/pipeline.md -- Current pipeline (what's already in motion)context/channels.md -- Lead sources, channel performancecontext/competitors.md -- Competitive landscapebrand/one-pagers/ -- Existing one-pagers (check before creating new)brand/brand-guide.md -- Brand colors, entity rulesBuild a complete go-to-market plan for a Labs product.
Trigger: "Launch strategy for [product]", "GTM plan for [product]", "How do we launch [product]?"
Execution:
GTM Plan Structure:
LAUNCH PLAN: [Product Name]
Date: [Date]
Phase Gate Status: [Pre-Launch / Soft Launch / Full Launch]
1. MARKET CONTEXT
- Target market size (mid-market DACH manufacturers, [segment])
- Current competitive alternatives
- Why now (market timing, regulatory drivers, technology shifts)
...Manage H2W Labs product roadmaps, PoC pipeline, customer feedback synthesis, product briefs, and conversion funnel analysis. Use when Lennart asks about Labs product priorities, feature backlog, PoC status, product feedback, conversion bottlenecks, or wants a product brief written.
Triggers: "Labs roadmap", "PoC pipeline", "product brief"
Data sources: Close, Jira, Moco, Hedy
Product management for H2W Labs: roadmap prioritization, PoC pipeline tracking, customer feedback analysis, product briefs, and conversion funnel diagnostics. Covers KnowKit, Product Assistant, EasyShare, Factory Track, and E-Invoice Integration.
Scope: H2W Labs products only. ERP consulting delivery uses separate project management processes.
Use this skill when Lennart says:
Always read before any mode:
context/offerings.md -- H2W Labs product catalog, pricing, positioningcontext/current-priorities.md -- Strategic priorities and time-sensitive itemscontext/pipeline.md -- Sales pipeline snapshot (Labs deals)context/strategy.md -- Growth thesis, Labs revenue targetsreferences/h2w-labs-crm-stages.md -- Close CRM stage mapping for Labs dealsLoad on demand:
context/clients.md -- Client profiles for customer-specific analysiscontext/deal-patterns.md -- Sales cycle benchmarks, conversion patternscontext/win-loss.md -- Deal outcome patternscontext/team-assignments.md -- Richard/Johannes availabilitylead_search / fetch_lead -- company records, contactsfind_opportunities -- Labs opportunities by stageactivity_search -- communications, notes, custom activitiessearch / paginate_search -- filtered deal queriesGetSessions -- find meetings with Labs prospects/customersGetSessionDetails -- full transcript for feedback extraction. Use raw transcript, never Hedy's recap.searchJiraIssuesUsingJql (cloudId: 0ee5c601-cc7e-4022-ba05-749fee885399) -- Labs project issues, PoC taskscreateJiraIssue -- create tickets for new features/tasks...
"Log daily time entries to Moco based on calendar events and manual input. Use when Lennart says log time, Zeiterfassung, log my hours, time tracking, update Moco time, or as part of session-closeout."
Triggers: "Log time", "Zeiterfassung", "log my hours". Interactive version with approval gate. Closeout skill handles autonomous posting.
Data sources: M365 calendar, Moco activities API, data/moco-time-mapping.json
Calendar-based daily time tracking. Pulls calendar events, maps them to Moco projects/tasks, fills gaps with manual input, and posts after approval.
Use when Lennart says:
/session-closeoutRead data/moco-time-mapping.json for project IDs, task IDs, and calendar patterns.
For each target date, pull calendar events via M365:
The calendar command in m365.sh only shows FUTURE events, missing past meetings. Instead, query the Graph API directly for the full day:
# Get access token
export $(grep MS_H2W_CLIENT_ID .env) && export $(grep MS_H2W_CLIENT_SECRET .env) && export $(grep MS_H2W_REFRESH_TOKEN .env) && export $(grep MS_H2W_TENANT_ID .env)
TOKEN=$(curl -s -X POST "https://login.microsoftonline.com/$MS_H2W_TENANT_ID/oauth2/v2.0/token" \
-d "client_id=$MS_H2W_CLIENT_ID" -d "client_secret=$MS_H2W_CLIENT_SECRET" \
-d "refresh_token=$MS_H2W_REFRESH_TOKEN" -d "grant_type=refresh_token" \
-d "scope=https://graph.microsoft.com/.default" | python -c "import json,sys; print(json.load(sys.stdin).get('access_token',''))")
# Query full day
curl -s -H "Authorization: Bearer $TOKEN" \
-H 'Prefer: outlook.timezone="Eastern Standard Time"' \
"https://graph.microsoft.com/v1.0/me/calendarView?startDateTime=2026-03-30T00:00:00&endDateTime=2026-03-30T23:59:59&\$top=50&\$select=subject,start,end,location,attendees,isAllDay&\$orderby=start/dateTime"Parse events for: subject, start time, end time, attendees (check for external domains), isAllDay. Skip all-day events (reminders, not time entries).
For each calendar event:
calendar_patterns in the mapping config (regex, case-insensitive). First match wins.sales_meetings...
Size markets, analyze segments, and assess expansion opportunities for ERP consulting and AI products. Use when Lennart says market size, TAM SAM SOM, market analysis, addressable market, market opportunity, how big is the market, segment analysis, should we enter [market], or when a proposal or strategic review needs market context.
Triggers: "Market size", "TAM SAM SOM", "addressable market"
Data sources: Apollo, Web, context files
Calculates TAM/SAM/SOM for specific market segments, analyzes customer segments in depth, and assesses the commercial viability of entering new markets. Designed for a consulting business (not SaaS), so all sizing is based on services spend, not license revenue.
This skill feeds into strategic-review, cso-analysis, labs-launch-strategy, and proposal-sow-generation. It can also produce a standalone one-page market sizing summary for inclusion in proposals or board-level documents.
Full TAM/SAM/SOM calculation for a specific market segment.
Examples: ERP consulting in DACH, AI applications for manufacturing in Germany, Infor LN services globally.
Method: Triple triangulation
The final estimate is the range where at least two of the three methods overlap.
TAM/SAM/SOM definitions (consulting-adapted):
| Level | Definition | Example |
|---|---|---|
| TAM | Total annual spend on this service category in the broadest relevant market | All ERP consulting spend in DACH: EUR 4-6B |
| SAM | The slice we could realistically serve given our platform focus, geography, and ICP | Infor LN consulting for mid-market manufacturers in DACH: EUR 200-400M |
| SOM | What we can realistically capture in 3 years given team size, brand, and pipeline | EUR 2-6M (0.5-2% of SAM, typical for a 15-person firm) |
...
Turn project experiences, customer conversations, and operational insights into content across multiple formats and channels. Use when Lennart asks to write a LinkedIn post, create a case study, draft an article, generate content ideas, repurpose content, write a video script, or produce any marketing or thought leadership content.
Triggers: "Write LinkedIn post", "create case study"
Data sources: Context files, content queue
Turn project experiences, customer conversations, and operational insights into content across multiple formats and channels.
Use this skill when Lennart says:
...
Parallel meeting prep engine. Spawns 4 subagents (Close CRM, Jira, Moco, vault+web) simultaneously, merges outputs, computes relationship health score, and produces an extended briefing note as markdown and one-page PDF. Use when Lennart asks to research a company, prep for a meeting, or gather background on a prospect/client.
Triggers: "Prep for [client]", "research [company]", "meeting prep"
Data sources: Close, Jira, Moco, M365, Hedy, Web
Parallel data-gathering system for pre-meeting intelligence. Spawns 4 subagents simultaneously, merges results, scores relationship health, and outputs both markdown and PDF.
Use this skill when Lennart says:
Before launching agents, determine:
If the meeting is today or tomorrow, pull calendar context first:
bash scripts/m365.sh calendar 7Use --account blc if the meeting is on the BLC calendar.
Launch ALL FOUR agents in a single message using the Agent tool. Each agent gets a targeted research brief.
Research brief for [CLIENT NAME]:
1. lead_search for "[client name]" -- find the lead
2. fetch_lead for the lead ID -- full details (contacts, custom fields, tags, notes)
3. find_opportunities for this lead -- ALL statuses (active, won, lost)
4. activity_search for this lead -- last 90 days, all activity types
5. For each contact on the lead: note name, title, email, last activity date
Return: lead details, all contacts with titles, all opportunities with stage/value/confidence/owner, last 5 activities with dates and summaries, days since last activity.Research brief for [CLIENT NAME]:
Using cloudId "0ee5c601-cc7e-4022-ba05-749fee885399":
1. searchJiraIssuesUsingJql: summary ~ "[client name]" OR labels = "[client name]" AND status != Done ORDER BY updated DESC (max 20 results)
2. searchJiraIssuesUsingJql: summary ~ "[client name]" AND (status = Blocked OR status = "On Hold") ORDER BY priority DESC
3. searchJiraIssuesUsingJql: summary ~ "[client name]" AND due <= now() AND status != Done ORDER BY due ASC
...Full meeting lifecycle -- preparation, notes processing, and follow-up drafting. Use when Lennart mentions meeting prep, meeting notes, meeting transcript, follow-up emails, post-meeting actions, or asks to prepare for or debrief after any meeting.
Triggers: "Meeting prep", "process meeting notes"
Data sources: Hedy, context files
Full meeting lifecycle: preparation, notes processing, and follow-up drafting.
Use this skill when Lennart says:
bash scripts/m365.sh calendar 7 -- pull upcoming meetings for contextbash scripts/m365.sh mail-search "[company or person name]" -- find recent email threads with the contactbash scripts/m365.sh mail-draft -- create follow-up email as Outlook draft (Phase 3)--account blc for BLC entity correspondencelead_search or search -- relationship history, open opportunitiesactivity_search -- recent CRM activity with the contactGetSessions -- find past meetings with this contactGetSessionDetails -- get full transcript (transcript or cleaned_transcript). Always use the raw transcript as primary input, not Hedy's recap or meeting_minutes.When asked to prepare for a meeting, delegate to the meeting-briefing-notes skill.
That skill runs 4 parallel subagents (Close CRM, Jira, Moco, vault+web), merges their outputs, computes a relationship health score, and produces both a markdown briefing and a one-page PDF.
reports/internal/meeting-prep/YYYY-MM-DD-{client-slug}.pdfIf invoked as part of meeting-workflow, proceed to Phase 2 after the meeting happens.
...
Cross-check Moco projects/deals against Close CRM opportunities. Flag misalignments, missing entries, and value discrepancies.
Triggers: "Sync check", "deal alignment", "moco crm sync"
Data sources: Moco, Close
Automated alignment check between Moco (system of record for projects) and Close CRM (system of record for pipeline).
moco crm sync, check alignment, sync check, deal alignment, project alignment, are moco and crm in sync
Launch two agents in parallel:
Moco agent:
export $(grep MOCO_API_TOKEN .env)
# All active projects with deals
curl -s -H "Authorization: Token token=$MOCO_API_TOKEN" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/projects" | \
python -c "import sys,json; data=json.load(sys.stdin); [print(json.dumps({'id':p['id'],'name':p['name'],'customer':p['customer']['name'],'active':p['active'],'budget':p.get('budget',0),'deal_id':p.get('deal_id')})) for p in data]"
# All deals
curl -s -H "Authorization: Token token=$MOCO_API_TOKEN" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/deals" | \
python -c "import sys,json; data=json.load(sys.stdin); [print(json.dumps({'id':d['id'],'name':d['name'],'company':d.get('company',{}).get('name',''),'status':d['status'],'value':d.get('value',0),'currency':d.get('currency','')})) for d in data]"CRM agent: Use find_opportunities (paginate through all) to get: opportunity ID, name, value, status, lead name, stage.
Check for these misalignment categories:
Exception: Projects under umbrella CRM opps (ARI, IMA, Neudorff W-xx). Check if parent customer has an umbrella opp covering the project.
...
Generate the Monday morning BD-PMO weekly update call briefing. Pulls from Close CRM, Jira, Hedy, Moco, and Slack to prepare agenda talking points. Use when Lennart says Monday briefing, prep the Monday meeting, BD-PMO call, weekly update call, or Monday morning report.
Triggers: "Monday briefing", "prep the Monday meeting"
Data sources: Close, Jira, Moco, Hedy, Slack
Generates the pre-meeting briefing report for the Monday morning Business Development and PMO weekly update call (7 AM EST). Pulls data from Close CRM, Jira, Hedy, Moco, and Slack, then organizes it into the meeting's agenda structure so each presenter has their numbers and talking points ready.
Use this skill whenever Lennart mentions the Monday meeting, BD-PMO call, Monday briefing, weekly update call, Monday morning report, or asks to prep the Monday meeting. Also trigger when he asks about marketing KPIs, sales KPIs, PMO KPIs, or utilization rates in the context of a meeting prep.
Use this skill when Lennart says:
find_opportunities -- active pipeline, new/won/lost dealsactivity_search -- emails, calls, meetings, notes (last 7 days)aggregation -- activity counts by type and userfind_pipelines_and_opportunity_statuses -- pipeline stagesorg_users -- map user IDs to namessearchJiraIssuesUsingJql with cloudId 0ee5c601-cc7e-4022-ba05-749fee885399getJiraIssue -- issue detailsGetSessions -- meetings from last 7 daysGetSessionDetails -- recaps, action items, highlightsGetToDos -- outstanding to-dos/projects -- active projects and budgets/users -- team members/schedules -- planned hours/activities -- logged hoursbash scripts/m365.sh calendar 7 -- upcoming meetings this week (both H2W and BLC accounts)bash scripts/m365.sh inbox 10 -- recent emails for context--account blc to also pull the BLC calendar...
Run the full Monday morning report sequence in one command. Chains context-refresh, c-suite-dashboard, monday-briefing, and weekly-snapshot in the correct order with consistent data. The c-suite-dashboard replaces the former top-management-report with a full C-level executive review. Use when Lennart says Monday sequence, run the Monday reports, full Monday, Monday morning, or start the week.
Triggers: "Monday sequence", "run the Monday reports"
Data sources: All
Runs the complete Monday morning report pipeline in one pass. Each step builds on the previous one so numbers are consistent across all reports.
Use this skill when Lennart says:
Check the "Last updated" dates in context/pipeline.md and context/projects.md.
This ensures all downstream reports pull from fresh context.
Run the deal-health-scoring skill. Scores all ~90 qualified-funnel opps and writes Tier + Score back to Close.
python scripts/deal_health_score.pyReads the generated data/deal-health/YYYY-MM-DD.json and captures the Red/Stalled/Yellow counts for the Monday briefing. If the JSON already exists for today (previous run in the same session), skip re-scoring.
Run the c-suite-dashboard skill.
This replaces the former top-management-report. The C-suite dashboard pulls all data (Qonto, Close CRM, Moco, Jira, Hedy, M365) in parallel via subagents and generates:
Output: Combined C-suite dashboard + individual role reports.
Important: Capture the key numbers from the dashboard (cash balances, pipeline totals, utilization %, team alerts, strategic status) in working memory. The monday-briefing must reference these exact numbers for consistency.
Run the monday-briefing skill.
Key rule: Reference the numbers from the C-suite dashboard (Step 1) rather than re-pulling Qonto and Close CRM. This guarantees consistency between the executive dashboard and the company briefing. The monday-briefing summarizes C-suite dashboard highlights into BD-PMO call talking points.
Additional data pulled:
...
Monthly bilingual newsletter compilation from published content. Curates top pieces, creates Close CRM email templates and workflows for sending. Use when Lennart says assemble newsletter, monthly newsletter, prepare the newsletter, or newsletter draft.
Triggers: "Assemble newsletter", "monthly newsletter"
Data sources: Published content, Close CRM
Compiles the monthly newsletter from the best-performing published content. Produces separate German (DACH) and English (international) editions. Sent via Close CRM workflows, following the existing VIP pattern (separate workflows per gender variant for German, single workflow for English).
Use this skill when Lennart says:
Also proactively suggest:
Always read before assembly:
data/content-queue/published/ -- Published content with performance datadata/content-queue/queue.md -- Queue for any newsletter-specific itemscontext/marketing.md -- Newsletter targets, campaign ideascontext/offerings.md -- Current product/service details for CTAscontext/current-priorities.md -- Strategic focus for exclusive insight sectionEach edition (DE and EN) follows this structure:
Newsletter delivery uses Close CRM workflows, following the proven VIP pattern.
German Edition (DACH) -- 3 workflow variants:
...
"Prepare lightweight briefings for recurring 1:1 meetings with team members. Pulls last meeting notes, recent Jira activity, Moco hours, and open action items. Use when Lennart asks to prep for a 1:1 with Uli, Richard, Lucia, Heiko, Diego, or any team member, or says 1:1 prep, catch me up on [person], or what did we discuss last time."
Triggers: "Prep for 1:1 with [person]", "catch me up on [person]"
Data sources: Hedy, Jira, Moco
Lightweight briefing for recurring internal 1:1 meetings. Unlike meeting-briefing-notes (designed for external company research with Apollo enrichment), this skill focuses on continuity: what happened since last time, what's on their plate, and what to discuss.
Use this skill when Lennart says:
| Person | Frequency | Focus Areas |
|---|---|---|
| Uli Schneider | Weekly | Project updates, Neudorff status, redeployment |
| Richard | As needed | Strategy, operations, technical decisions |
| Lucia Morandini | Daily | HR, admin, onboarding, office management |
| Heiko | Via BD-PMO Monday | PM coordination, delivery, resource allocation |
| Diego | Via BD-PMO Monday | PM coordination, project status |
GetSessions -- find last 1:1 with this personGetSessionDetails -- get full transcript (transcript or cleaned_transcript), then extract action items, decisions, and discussion topics from the raw conversation. Do not rely on Hedy's recap or meeting_minutes.searchJiraIssuesUsingJql -- issues assigned to or updated by this person (last 14 days)0ee5c601-cc7e-4022-ba05-749fee885399GET /activities?user_id={id}&from=YYYY-MM-DD -- recent time entriesGET /planning_entries?user_id={id} -- upcoming assignmentsAuthorization: Token token=$MOCO_API_KEYcontext/team.md -- role, responsibilitiescontext/team-assignments.md -- current project assignmentscontext/team.md to identify the person's role and responsibilitiescontext/team-assignments.md for their current project assignmentsHedy (main conversation):
GetSessions with limit 10 -- scan titles for this person's nameGetSessionDetails for the most recent match -- get full transcript and extract action items, decisions, key topics from the raw conversationJira (jira-agent or direct):
...
Operational health monitoring across utilization, delivery, pipeline, and finance. Use when Lennart asks for an ops dashboard, weekly ops check, KPI review, how the business is running, delivery health, or wants a cross-functional status view. Pulls from Moco, Jira, Close CRM, and Qonto to give one unified operational picture.
Triggers: "Ops dashboard", "KPI review"
Data sources: Moco, Jira, Close, Qonto
Cross-functional operations dashboard that pulls from all systems to give a unified view of how the business is actually running. Goes beyond individual skill reports (utilization, pipeline velocity, cash position) by connecting the dots across functions and flagging where one area's problem affects another.
Use this skill when Lennart says:
Always read before analysis:
context/kpis.md -- KPI targets and thresholds (GREEN/YELLOW/RED). Includes internal cost ratio KPI (target <30%, historical: 53% in 2025).context/finance.md -- Cost structure, billing rates, margin targets, actual rate realization, internal cost ratio trendcontext/team-capabilities.md -- Team capacity, key-person risks, historical utilization baselines per personcontext/team-assignments.md -- Who is on whatcontext/current-priorities.md -- Strategic contextcontext/risks.md -- Active business risksWhen scoring operational health, include internal cost ratio as a KPI:
Report endpoints (use FIRST, pre-calculated):
GET /report/utilization?from=&to= -- daily billable/unbillable per user WITH target hours. Eliminates manual ratio math.GET /report/planned_vs_tracked?from=&to= -- planned vs actual hours by user+project with deltaGET /report/absences?year=YYYY&active=true -- vacation/sick days per userGET /projects/{id}/report -- budget, hours, costs, profitability per project (single call)GET /account/fixed_costs?year=YYYY -- monthly fixed costs by categoryRaw data endpoints:
GET /users -- team rosterGET /activities?from=YYYY-MM-DD&to=YYYY-MM-DD -- time entries (use for project-level breakdown)GET /projects?include_archived=false -- active projects, budgets, timelinesGET /invoices -- invoicing data...
"Pull a comprehensive briefing on any team member by combining their people record with fresh Moco, email, Jira, and Hedy data. Use when Lennart says prep for [name], catch me up on [name], what's [name] working on, or before any 1:1 or evaluation meeting."
Triggers: "Prep for [name]", "catch me up on [name]"
Data sources: Moco, M365, Jira, Hedy, data/people/
Pull a comprehensive briefing on a team member by combining their persistent people record (data/people/) with fresh data from Moco, M365, Jira, and Hedy.
Use this skill when Lennart says:
data/people/*.md and find the file matching the person's name (case-insensitive partial match on filename)moco_id from YAML frontmatter for API callsMoco (performance data, pre-calculated):
export $(grep MOCO_API_KEY .env)
# Use performance_report for pre-calculated annual+monthly breakdown
curl -s -H "Authorization: Token token=$MOCO_API_KEY" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/users/{moco_id}/performance_report?year=2026"
# Returns: target_hours, hours_tracked_total, hours_billable_total, variation per month
# For project-level detail, also pull activities
curl -s -H "Authorization: Token token=$MOCO_API_KEY" \
"https://h2wconsultinggmbh.mocoapp.com/api/v1/activities?user_id={moco_id}&from={30_days_ago}&to={today}"Parse: performance_report for target vs actual (pre-calculated), activities for project-level breakdown.
M365 (recent emails, last 14 days):
bash scripts/m365.sh mail-search "from:{email}" 10 --account h2w
bash scripts/m365.sh mail-search "to:{email}" 10 --account h2wParse: recent email subjects, who they're communicating with, any client-facing threads.
Jira (assigned issues): Use searchJiraIssuesUsingJql:
assignee = "{name}" ORDER BY updated DESC (last 20)0ee5c601-cc7e-4022-ba05-749fee885399Hedy (recent meetings with this person): Use GetSessions limit 20, scan titles for person's name or company names from their accounts. For matches, GetSessionDetails to get the full transcript (transcript or cleaned_transcript). Extract action items and discussion topics from the raw conversation, not from Hedy's recap or summary.
Issue Tracker (open issues involving this person):
...
Weekly pipeline velocity report from Close CRM. Shows deals by stage, age in stage, stalled deals, conversion rates, weighted revenue, and win/loss analysis. Use when Lennart asks about pipeline velocity, deal flow, stalled deals, conversion rates, win/loss, or how fast deals are moving.
Triggers: "Pipeline velocity", "stalled deals"
Data sources: Close
Measures how fast deals move through the pipeline, where they stall, what converts, and what falls out. Uses Close CRM as the single data source. This is the sales rhythm report that drives weekly pipeline reviews.
Use this skill when Lennart says:
Triggered by: "pipeline velocity", "pipeline review", "deal flow" Shows: Full pipeline snapshot with movement, aging, and velocity metrics.
Triggered by: "stalled deals", "what's stuck", "dead deals" Shows: Deals that haven't moved stages or had activity in 14+ days, ranked by value.
Triggered by: "win/loss", "what did we win", "conversion rates" Shows: Deals won and lost in the last 30/60/90 days with reasons, cycle times, and patterns.
Triggered by: "proposal tracker", "what proposals are out", "pending proposals" Shows: Deals in proposal/negotiation stage with age, value, and follow-up status.
find_opportunities -- deals by status_type (active, won, lost), with filtersfind_pipelines_and_opportunity_statuses -- pipeline structure and stage definitionsactivity_search -- recent activity on deals (emails, calls, notes)aggregation -- activity counts grouped by type/userorg_users -- map user IDs to namessearch / lead_search -- lead details for deal contextPhase 1: Pipeline Data (research-agent)
Launch via Agent tool with subagent_type: "research-agent". Brief:
find_pipelines_and_opportunity_statuses -- get all pipeline stages with IDsfind_opportunities with status_type "active" -- full active pipelinefind_opportunities with status_type "won" and date_updated >= 7 days agofind_opportunities with status_type "lost" and date_updated >= 7 days agoorg_users -- map user IDs to names...
Generate and edit PowerPoint presentations from templates. Use when user needs to create slides from outlines, modify existing decks, profile custom templates, or validate presentation quality.
Triggers: Engine: outline to PPTX, edit decks, validate, profile templates
Data sources: Brand template, config
Generate consultant-ready PowerPoint presentations from markdown outlines.
# Generate from outline (Inner Chapter template)
python .claude/skills/pptx-from-layouts/scripts/generate.py outline.md -o deck.pptx
# Edit existing deck
python .claude/skills/pptx-from-layouts/scripts/edit.py deck.pptx --inventory
python .claude/skills/pptx-from-layouts/scripts/edit.py deck.pptx --replace '{"slide":3,"old":"2025","new":"2026"}'
# Validate quality
python .claude/skills/pptx-from-layouts/scripts/validate.py deck.pptx
# Profile custom template
python .claude/skills/pptx-from-layouts/scripts/profile.py template.pptx --generate-config# Basic generation
python .claude/skills/pptx-from-layouts/scripts/generate.py outline.md -o output.pptx
# With validation
python .claude/skills/pptx-from-layouts/scripts/generate.py outline.md -o output.pptx --validate
# Custom template
python .claude/skills/pptx-from-layouts/scripts/generate.py outline.md -o output.pptx \
--config custom-config.json --template custom-template.pptx
# Parse only (no PPTX)
python .claude/skills/pptx-from-layouts/scripts/generate.py outline.md --layout-only -o layout.jsonUse for text-only changes to < 30% of slides.
# Extract content inventory
python .claude/skills/pptx-from-layouts/scripts/edit.py deck.pptx --inventory
# Replace text (inline JSON)
python .claude/skills/pptx-from-layouts/scripts/edit.py deck.pptx \
--replace '{"slide":3,"old":"Q1 2025","new":"Q2 2026"}'
# Replace from file
python .claude/skills/pptx-from-layouts/scripts/edit.py deck.pptx --replace changes.json -o edited.pptx
# Reorder slides (0-indexed)
python .claude/skills/pptx-from-layouts/scripts/edit.py deck.pptx --reorder "0,2,1,3,4" -o reordered.pptx# Basic quality check
python .claude/skills/pptx-from-layouts/scripts/validate.py deck.pptx
# With layout coverage analysis
python .claude/skills/pptx-from-layouts/scripts/validate.py deck.pptx --template template.pptx
# Compare to reference
python .claude/skills/pptx-from-layouts/scripts/validate.py deck.pptx --reference expected.pptx
# Generate diff report
...Create on-brand PPTX presentations for any topic. Use when user asks to create a presentation, slide deck, pitch deck, or PPTX file.
Triggers: "Create a deck", "pitch deck", "slide deck", "PPTX"
Data sources: Brand template, config
Create professional PowerPoint presentations in H2W/BLCG/H2W Labs brand style.
Always ask these three questions:
Every presentation follows a structure based on audience type. These are non-negotiable.
Based on actual H2W/BLCG client decks (MeisterWerke, W. Neudorff).
Slide 1: Cover (client name, title, subtitle, date, entity logos)
Slide 2: Disclaimer (standard copyright/IP text)
Slide 3: Agenda (numbered sections matching content)
Slide 4+: Content slides (with copyright footer on each)
Use section dividers between major topics
Slide N: Contact / Closing (entity addresses, websites, copyright)Based on the H2W Labs company introduction deck. Problem-led narrative, no disclaimer or agenda.
Slide 1: Cover (H2W Labs branding, title, subtitle)
Slide 2: Problem statement (why this matters)
Slide 3: Current situation (why AI initiatives fail / pain points)
Slide 4: H2W Labs positioning (who we are, why we exist)
Slide 5+: Services, workshops, products, architecture
Slide N: Closing ("Vielen Dank fuer Ihre Aufmerksamkeit!" + contact)Lighter structure. No disclaimer or formal agenda required.
Slide 1: Cover (title, date, context)
Slide 2+: Content slides
Slide N: Closing or summaryUse on slide 2 of every client-facing proposal/positioning deck:
Disclaimer
Copyright (c) {year} H2W Consulting GmbH, BL Consulting Group LLC,
BL Consulting UG (haftungsbeschraenkt) und verbundene Unternehmen.
Alle Rechte vorbehalten.
Diese Praesentation und ihr Inhalt sind geistiges Eigentum der
H2W Consulting GmbH und der BL Consulting Group. Die Weitergabe,
Vervielfaeltigung oder Verwendung, auch auszugsweise, ist ohne
vorherige schriftliche Genehmigung nicht gestattet.
Die in dieser Praesentation enthaltenen Informationen sind vertraulich
...Validate, repair, and audit PowerPoint presentations. Use when polishing a generated deck, checking an existing PPTX for quality issues, or running QA before delivery.
Triggers: QA: audit, repair, validate any PPTX. Text overflow, ghosts, fonts, hyperlinks, ZIP fixes
Data sources: python-pptx
Post-generation QA for any PowerPoint file. Validates, repairs, and produces a slide-by-slide audit report.
# Full pipeline: audit -> repair -> validate -> report
python .claude/skills/pptx-polish/scripts/polish.py deck.pptx
# Audit only (no fixes)
python .claude/skills/pptx-polish/scripts/polish.py deck.pptx --audit-only
# Fix only (no report)
python .claude/skills/pptx-polish/scripts/polish.py deck.pptx --fix-only
# Individual scripts
python .claude/skills/pptx-polish/scripts/audit.py deck.pptx
python .claude/skills/pptx-polish/scripts/repair.py deck.pptx
python .claude/skills/pptx-polish/scripts/validate.py deck.pptx| Check | Severity | Threshold |
|---|---|---|
| Text overflow (chars) | warning | >800 per shape |
| Text overflow (paragraphs) | warning | >15 per shape |
| Text overflow (line length) | warning | >120 chars per line |
| Ghost placeholders | warning | Empty with default text |
| Font consistency | warning | Mixed fonts per slide |
| Unresolved theme tokens | info | +mn-ea, +mj-lt in fonts |
| Empty slides | warning | No visible text content |
| Malformed hyperlinks | warning | Empty or invalid URLs |
| Bare URLs in text | info | URL without hyperlink |
| Missing title | info | No title shape |
| Repair | Description |
|---|---|
| ZIP dedup | Remove duplicate ZIP entries (LAST-wins) |
| SVG content type | Add missing image/svg+xml to Content_Types |
| Ghost placeholders | Remove empty placeholders with default text |
| Hyperlink linkification | Add hyperlinks to bare URLs in text |
| Dead hyperlink cleanup | Remove hyperlinks with empty targets |
{name}-audit.md with slide-by-slide findings and before/after comparisonChain after pptx-from-layouts or pptx-generator:
python .claude/skills/pptx-from-layouts/scripts/generate.py outline.md -o deck.pptx
python .claude/skills/pptx-polish/scripts/polish.py deck.pptxScaled content strategy design for programmatic SEO. Evaluates feasibility, designs template-based page strategies, and generates keyword x topic matrices for batch production. Use when Lennart says programmatic SEO, scaled content, content at scale, location pages, topic variants, keyword matrix, or template pages.
Triggers: "Programmatic SEO", "scaled content", "keyword matrix"
Data sources: Context files, seo-keywords
Designs and evaluates template-based content strategies that produce many pages from structured patterns. Where blog-writer handles individual articles and content-engine handles batch scheduling, this skill handles the strategic layer: deciding which page patterns to pursue, scoring their viability, and generating the content matrices that feed production.
Target domain: h2w-consulting.com (bilingual DE/EN).
Use this skill when Lennart says:
Always read before execution:
context/seo-keywords.md -- Keyword targets and ranking statuscontext/offerings.md -- Services and products (determines what pages we can credibly produce)context/marketing.md -- Content pillars, existing inventorycontext/icp.md -- Buyer personas and industry segmentscontext/clients.md -- Client base (informs which industries we have real experience in)context/strategy.md -- Growth thesis and market positionScore a proposed programmatic SEO pattern before investing in production.
Trigger: "Should we do [pattern] pages?", "Assess [pattern] feasibility", "Score this SEO idea"
Input: A page pattern (e.g., "[Service] for [Industry]" or "[City] ERP consulting")
Feasibility Scoring Framework (6 dimensions, 100 points total):
| Dimension | Weight | What It Measures | Scoring Guide |
|---|---|---|---|
| Search Pattern Validity | 20 | Do people actually search this pattern? | 0-5: No evidence. 6-10: Sporadic. 11-15: Consistent volume across variants. 16-20: High volume, clear pattern. |
| Unique Value Per Page | 25 | Can each page offer genuinely different content? | 0-8: Same content with swapped nouns. 9-16: Moderate differentiation. 17-25: Each page has unique insights, data, or recommendations. |
| Data Quality | 20 | Do we have proprietary or differentiated data to populate pages? | 0-7: Public data only. 8-14: Unique analysis of public data. 15-20: Proprietary experience, project data, client patterns. |
...
Regenerate and deploy the password-gated EA project showcase website at webapp/showcase/. Use when Lennart says regenerate showcase, update showcase, rebuild the showcase site, deploy showcase, project showcase, or showcase site.
Triggers: n/a
Data sources: n/a
Password-gated single-file website at webapp/showcase/ that documents every skill, subagent, rule, integration, workflow, recent activity, decision log excerpt, and gallery artifact in the EA project. Used as Lennart's personal memory aid and as a demo.
Target URL: https://ea.blconsultingservice.com (Netlify) Password: ea-showcase2026
| Mode | Trigger | Action |
|---|---|---|
| regenerate | "regenerate showcase", "update showcase", "rebuild showcase" | Run python generate_site.py in webapp/showcase/. Verify counts. |
| deploy | "deploy showcase", "push showcase" | Run generator + npx netlify deploy --prod --dir=. |
| edit narrative | "update showcase intro", "edit showcase philosophy", etc. | Edit webapp/showcase/content/*.md, then regenerate. |
The generator parses these sources on every build, so the catalog stays in sync with the project automatically:
.claude/skills/*/SKILL.md — all skills (frontmatter + body excerpt, searchable).claude/agents/*.md — subagents.claude/rules/*.md — behavioral rulesscripts/*.{py,sh} — integration scripts, grouped by source systemreferences/skill-catalog.md — category mappingdata/daily-log/*.md — recent activity feed (last 8 days)decisions/log.md — decision log quick indexreports/**/*.pdf — gallery of generated artifactscontext/*.md — context files indexmemory/MEMORY.md — persistent memory indexgit log — timeline of skill additions + commit statsHand-written narrative lives in webapp/showcase/content/01-06-*.md. Edit these when the architecture, workflows, or story need updating.
cd webapp/showcase
python generate_site.py
start index.htmlVerify:
find .claude/skills -name SKILL.md -maxdepth 2 | wc -lreports/decisions/log.mdcd webapp/showcase
python generate_site.py
npx netlify deploy --prod --dir=....
Generate weekly project status reports in client-facing and internal versions. Use when Lennart asks to write a status report, prepare a project update, review a PM's status report draft, or summarize project status for any stakeholder.
Triggers: "Status report for [client]"
Data sources: Jira, Moco
Generate weekly project status reports in two versions: client-facing and internal.
Use this skill when Lennart says:
Audience: Project sponsor, IT lead, project stakeholders
Template:
PROJECT STATUS REPORT
Project: [Name]
Client: [Company]
Report date: [Date]
Reporting period: [From -- To]
OVERALL STATUS: [GREEN / YELLOW / RED]
KEY PROGRESS THIS PERIOD
- [Accomplishment 1]
- [Accomplishment 2]
- [Accomplishment 3]
UPCOMING MILESTONES
| Milestone | Target date | Status |
|-----------|------------|--------|
| [Name] | [Date] | [On track / At risk / Delayed] |
RISKS AND BLOCKERS
| Risk | Impact | Mitigation | Owner |
|------|--------|------------|-------|
| [Description] | [High/Med/Low] | [Action] | [Person] |
ACTIONS REQUIRED FROM CUSTOMER
- [Specific action needed, with deadline]
NEXT STEPS
- [What happens next period]Client report rules:
Audience: Project manager, consulting team, Richard, Heiko, Lennart
Template:
INTERNAL PROJECT STATUS
Project: [Name]
Client: [Company]
Report date: [Date]
PM: [Name]
OVERALL STATUS: [GREEN / YELLOW / RED]
BUDGET VS. ACTUAL
| Phase / Stream | Planned effort | Actual effort | Remaining | Variance |
|---------------|---------------|--------------|-----------|----------|
| [Phase] | [X days] | [X days] | [X days] | [+/- X] |
| TOTAL | [X days] | [X days] | [X days] | [+/- X] |
RESOURCE ALLOCATION
| Consultant | Role | Planned days | Actual days | Next period |
...Generate consulting proposals and statements of work for ERP consulting and H2W Labs engagements. Use when Lennart asks to draft a proposal, write a SOW, put together an offer, create a quote, or prepare commercial documents for any client engagement.
Triggers: "Draft a proposal", "write a SOW"
Data sources: Context files
Generate consulting proposals and statements of work for ERP consulting and H2W Labs engagements.
Use this skill when Lennart says:
1. Executive Summary
2. Customer Situation
3. Proposed Approach
4. Scope of Services
5. Project Phases and Deliverables
6. Roles and Responsibilities
7. Timeline
8. Commercial Terms
9. Assumptions and Exclusions
Same structure, but adjust positioning:
| Engagement size | Target length |
|---|
...
Build and manage target account lists for systematic outbound prospecting. Uses Apollo.io to find ICP-matching companies, enriches with contact data, and stages outreach in Close CRM. Use when Lennart asks to build a prospect list, find target accounts, research potential clients, set up outbound campaigns, or identify companies that match our ICP.
Triggers: "Build prospect list", "find target accounts"
Data sources: Apollo, Close, ICP
Systematic outbound prospecting using Apollo.io for discovery, ICP scoring for qualification, and Close CRM for pipeline staging.
Use when Lennart says:
Read before executing:
context/icp.md -- ICP definitions, fit scoring (A/B/C/D), disqualification criteriacontext/offerings.md -- What we sell, pricing, deal archetypescontext/pipeline.md -- Existing pipeline to avoid duplicatescontext/clients.md -- Existing clients to avoid targetingcontext/channels.md -- Channel strategy by offeringFind companies matching ICP criteria. Input: segment/criteria. Output: scored prospect list.
Take leads already in Close CRM and enrich with Apollo.io data (contacts, revenue, tech stack, job postings).
Search for buying signals: Infor LN job postings, ERP-related hiring, digital transformation roles.
Deep research on a specific target account before outreach. Company profile, decision makers, triggers, recommended approach.
Map Lennart's request to Apollo.io search parameters:
| ICP Criterion | Apollo.io Parameter |
|---|---|
| Industry: Manufacturing | Industry keywords: manufacturing, industrial |
| Revenue: EUR 50-500M | Revenue range: $50M-$500M |
| Employees: 200-5000 | Employee count range |
| Geography: DACH | HQ country: Germany, Austria, Switzerland |
| ERP: Infor LN | Technology keywords (limited), job postings |
Use apollo_mixed_companies_search with mapped criteria.
For each result, score against context/icp.md fit matrix:
...
Scrape web sources, qualify companies against ICP, enrich via Apollo, and push to Close CRM with follow-up tasks. Use when Lennart says scrape and prospect, intake from URL, prospect from [source], run intake on, prospecting intake, or provides a URL to scrape for leads.
Triggers: n/a
Data sources: n/a
Reusable pipeline: scrape a web source, ICP-qualify companies, get approval, enrich, push to Close CRM, create tasks.
Use when Lennart says:
Read before executing:
context/icp.md -- ICP definitions, fit scoring, disqualification criteriacontext/pipeline.md -- Existing pipeline to avoid duplicatescontext/clients.md -- Existing clients to avoid targetingscripts/scraper_presets.json that matches the URLpython scripts/scraper.py extract <url> --preset <name>fetch mode first to understand the page, then determine how to get the full listdata/prospecting/YYYY-MM-DD-<source-slug>.jsonStaging file schema:
{
"source": "<url>",
"source_type": "<preset or custom>",
"scraped_at": "<ISO datetime>",
"pipeline_status": "stage_1_scraped",
"companies": [
{
"name": "<company name>",
"signal": "<job title or other signal>",
"location": "<city/country>",
"source_url": "<link to specific listing>",
"icp_score": null,
"icp_match": [],
"icp_reject": [],
"status": "pending",
"approved": null,
"enrichment": {"company": null, "contacts": []},
"crm_lead_id": null,
"crm_contact_ids": [],
"crm_task_id": null
}
]
}Deduplication at scrape time: When multiple records share the same company name, group them. Store the first signal (job title) and note the count. One company entry per unique company, not per job posting.
For each company, score against context/icp.md:
Scoring heuristics (based on company name + signal + location, before enrichment):
...
"Prepare quarterly business reviews with key clients. Generates QBR agenda, delivery summary, financials, expansion opportunities, and follow-up actions. Use when Lennart asks to prepare a QBR, quarterly review with a client, client business review, or review a customer relationship formally."
Triggers: "Prepare QBR for [client]", "quarterly review with [client]"
Data sources: Close, Jira, Moco, Hedy
Prepare and run quarterly business reviews with key clients. Pulls delivery data, surfaces contract dates, expansion opportunities, and relationship health. Generates a structured QBR package or client-facing PDF.
Use this skill when Lennart says:
Trigger: "Prepare QBR for [client]", "quarterly review with [client]" Generate full QBR prep package: agenda, delivery summary, financial overview, expansion analysis, risk assessment, talking points.
Trigger: "QBR deck for [client]", "QBR report for [client]" Generate a client-facing QBR summary document as PDF via report engine.
Trigger: "QBR follow-up for [client]", "process QBR notes for [client]" Process QBR meeting notes from Hedy and generate follow-up actions, CRM updates, and next-steps email.
searchJiraIssuesUsingJql -- issues by project/customer (cloudId: 0ee5c601-cc7e-4022-ba05-749fee885399)getJiraIssue -- issue details for key itemsGET /projects -- filter by customer identifierGET /activities -- hours logged by projectGET /invoices -- invoicing historyAuthorization: Token token=$MOCO_API_KEYhttps://h2wconsultinggmbh.mocoapp.com/api/v1lead_search / fetch_lead -- customer record, contactsfind_opportunities -- open and closed dealsactivity_search -- recent activity (emails, calls, meetings)GetSessions -- find recent meetings with this clientGetSessionDetails -- get full transcript (transcript or cleaned_transcript). Extract decisions, action items, and discussion themes from the raw conversation. Do not rely on Hedy's recap or meeting_minutes.context/clients.md -- client profile, tier, lifetime valuecontext/contracts.md -- renewal dates, termscontext/revenue-transitions.md -- upcoming revenue changestemplates/quarterly-business-review.md -- agenda structure...
> Relationship intelligence dashboard. Monitors interaction frequency with key contacts and alerts when relationships need attention. Use when Lennart says relationship pulse, check my relationships, who needs attention, relationship dashboard, track a contact, or relationship detail.
Triggers: "Relationship pulse", "who needs attention", "check my relationships"
Data sources: M365, Close, Hedy, data/relationships/
Monitors interaction frequency with key contacts against their configured cadence targets and alerts when relationships are going cold. Combines M365 email, Close CRM activity, calendar, and Hedy meeting data into one prioritized view.
The registry lives in data/relationships/*.md. Each file defines who to track, at what cadence, and what role they play. This skill reads those files, scans live data sources, and reports who needs attention today.
Use this skill when Lennart says:
Also proactively suggest this skill:
Parse the mode from the user message:
| Mode | Trigger phrases | Default |
|---|---|---|
| scan | "pulse", "who needs attention", "dashboard", no qualifier | Yes |
| add | "add [name]", "track [name]", "start tracking" | No |
| remove | "remove [name]", "stop tracking [name]", "untrack" | No |
| list | "list contacts", "show all tracked", "who am I tracking" | No |
| detail | "detail [name]", "deep dive [name]", "[name] status" | No |
Read all files in data/relationships/:
README.md and any file where frontmatter paused: trueFor each contact, resolve cadence_days and extended_days from the contact's role and cadence field:
Cadence keyword to days:
Role defaults (used when cadence is not specified or for extended threshold):
| Role | cadence_days | extended_days |
|---|---|---|
| champion | 14 | 28 |
| decision-maker | 30 | 45 |
| gatekeeper | 30 | 45 |
| technical | 30 | 60 |
| partner | 30 | 45 |
...
Process a sales discovery meeting into actionable CRM updates, deal qualification, and follow-up emails. Pulls the Hedy transcript, extracts qualifying information (pain points, budget, timeline, decision makers, tech environment), updates Close CRM, recommends pipeline stage, and drafts the follow-up email. Use when Lennart says "debrief the call", "process the meeting with [company]", "update CRM from the [company] call", "what came out of the meeting", "debrief [company]", "post-meeting update", or after any sales discovery conversation. This skill bridges the gap between a meeting happening and the pipeline being updated -- it turns a conversation into structured deal intelligence and next actions.
Triggers: "Debrief the call", "process the meeting"
Data sources: Hedy, Close
Turns a sales discovery meeting transcript into structured deal intelligence, CRM updates, and a follow-up email. This is the bridge between "we had a meeting" and "the pipeline reflects reality."
Without this skill, post-meeting processing is manual: Lennart has to recall what was discussed, open Close CRM, update fields, figure out the right stage, draft a follow-up, and update internal tracking. This skill does all of that in one pass.
After any sales-relevant meeting -- discovery calls, scoping sessions, introductory conversations, product demos, or qualification calls. Works for both ERP consulting prospects and H2W Labs/AI prospects.
Typical triggers:
If Lennart specifies a company name or session, use that. Otherwise, use index-first lookup:
data/hedy-sessions-index.json and filter by customer name (fuzzy match) and recent date rangeGetSessionDetailsdata/hedy-topics-map.json: try GetTopicSessions(topicId) as alternativeGetSessions with limit 10, present list, ask which to processGetSessionDetails with the session IDKey fields to extract from the Hedy response:
transcript or cleaned_transcript (the raw conversation -- often in German)meeting_minutes (structured summary with action items)recap (concise summary with key points)user_todos (action items assigned to Lennart)The transcript is the richest source. The recap and meeting minutes are useful cross-references but may miss nuance. When in doubt, go to the transcript.
Parse the meeting content and extract these fields. Transcripts are often in German -- extract the information and present it in whatever language matches the client relationship (German for DACH clients, English for international).
Deal Qualification Card:
| Field | What to Look For |
|---|---|
| Company | Full legal name, location, industry |
...
Sales cockpit that pulls data from Close CRM, cross-references with Jira delivery, enriches with Apollo.io and web research. Use when Lennart asks about the pipeline, deals needing attention, a specific deal or company status, weekly sales briefing, prospecting research, revenue outlook, or any sales-related analysis.
Triggers: "Pipeline review", "deal intelligence"
Data sources: Close, Jira, Apollo, Web
Comprehensive sales cockpit that pulls data from Close CRM, cross-references with Jira project delivery, enriches with Apollo.io and web research, and delivers actionable intelligence with specific next-step recommendations.
This skill understands the business context (priorities, projects, goals) and connects the dots between sales pipeline, delivery status, and market intelligence.
Use this skill when Lennart says:
All tools are connected via MCP through Claude.ai. No API keys needed.
search / lead_search / paginate_search -- find leads and contactsfind_opportunities / fetch_opportunity -- pipeline and deal datafetch_lead -- lead details, contacts, custom fieldsactivity_search -- emails, calls, tasks, meetings, notesaggregation -- metrics (sum, average, count, group by field, time bucketing)find_pipelines_and_opportunity_statuses -- pipeline stagesorg_users -- team membersget_fields -- custom field definitionsfind_lead_smart_views -- saved viewssearchJiraIssuesUsingJql -- find project issues by customergetJiraIssue -- issue detailsgetVisibleJiraProjects -- active projectssearchConfluenceUsingCql / getConfluencePage -- project documentationsearch (Rovo) -- cross-system searchapollo_mixed_companies_search -- find companies by criteriaapollo_mixed_people_api_search -- find people by criteriaapollo_organizations_enrich -- detailed company dataapollo_people_match -- enrich specific contactsapollo_organizations_job_postings -- hiring signalsWebSearch -- company news, press releases, ERP/transformation signalsWebFetch -- pull specific pages for deeper context...
Draft outreach messages for sales scenarios across email and LinkedIn. Use when Lennart asks to write an outreach email, draft a LinkedIn message, follow up with a contact, re-engage a dormant lead, respond to an inbound inquiry, or craft any sales-related communication.
Triggers: "Write outreach email", "follow up with..."
Data sources: Context files
Draft outreach messages for different sales scenarios across email and LinkedIn.
Use this skill when Lennart says:
Typical personas at mid-sized industrial manufacturing companies:
All messaging should resonate with operational leaders who care about efficiency, reliability, and measurable results. Avoid generic tech/SaaS language.
Purpose: Introduce BLCG / H2W Consulting / H2W Labs to a company that fits our ideal customer profile.
Email structure:
LinkedIn structure:
Purpose: Continue a conversation started at a conference, webinar, or through a referral.
Structure:
Purpose: Restart a conversation with a former customer or cold lead.
Structure:
Purpose: Reply to someone who reached out to us.
Structure:
...
SEO auditing for h2w-consulting.com. Site health, content scoring, keyword strategy, and competitor SEO analysis. Use when Lennart says SEO audit, content audit, keyword review, SEO health, competitor SEO, or ranking check.
Triggers: "SEO audit", "content audit", "keyword review"
Data sources: Web, context files
Audits h2w-consulting.com for technical SEO health, content effectiveness, keyword strategy, and competitive positioning. Produces actionable reports with prioritized recommendations.
Use this skill when Lennart says:
Also proactively suggest:
Always read before execution:
context/seo-keywords.md -- Current keyword strategy and trackingcontext/marketing.md -- Content pillars, publishing cadence, inventorycontext/offerings.md -- Service/product catalog (for keyword-to-offering alignment)context/icp.md -- Target audience (search intent alignment)context/competitors.md -- Known competitors for Mode 3Weighted scoring system for site-level SEO health (0-100):
| Category | Weight | What It Measures |
|---|---|---|
| Crawlability | 30% | robots.txt, sitemap.xml, indexation, canonical tags, crawl errors |
| Technical | 25% | Page speed, mobile readiness, HTTPS, Core Web Vitals indicators, structured data |
| On-Page | 20% | Title tags, meta descriptions, heading hierarchy, URL structure, image alt text |
| Content | 15% | Keyword targeting, content depth, freshness, internal linking density |
| Authority | 10% | Backlink indicators, domain signals, brand search presence |
Score thresholds:
Per-article scoring template (0-100):
| Factor | Weight | Criteria |
|---|---|---|
| Keyword Targeting | 25% | Primary keyword in title, H1, first 100 words, meta description, URL slug |
...
"DEPRECATED -- use /closeout instead. Redirects to the closeout skill."
Triggers: n/a
Data sources: n/a
This skill has been replaced by /closeout, which enforces hard-sequenced steps with a completion checklist.
Run /closeout instead.
Optimize the prompts, structure, and efficiency of existing skills and agents. Use when Lennart says optimize skill, improve skill, skill quality, prompt optimization, reduce token usage, skill review, make this skill better, audit skills, tune agent, or fix skill output.
Triggers: "Optimize skill", "skill review", "prompt optimization"
Data sources: .claude/skills/, .claude/agents/
Meta-skill that reviews and improves existing skills and agent definitions for quality, clarity, efficiency, and output consistency. Where skill-pattern-detector finds gaps and recommends new skills, this skill fixes and improves the ones that already exist.
Use this skill when Lennart says:
Also proactively suggest when:
Review a single skill for quality across 5 dimensions. Run when asked to review or audit a specific skill.
Steps:
.claude/skills/{skill-name}/SKILL.mdreferences/skill-catalog.md for cross-referenceOutput: Skill Audit Report
# Skill Audit: {skill-name}
## Scores (X.X / 5.0 each)
| Dimension | Score | Key Issue |
|-----------|-------|-----------|
| Trigger Clarity | X.X | [one-line summary] |
| Instruction Precision | X.X | [one-line summary] |
| Integration Depth | X.X | [one-line summary] |
| Output Consistency | X.X | [one-line summary] |
| Token Efficiency | X.X | [one-line summary] |
| **Overall** | **X.X** | |
## Issues Found
### [Issue 1: title]
- **Dimension:** [which scoring dimension]
- **Location:** [line or section reference]
- **Problem:** [what is wrong]
- **Fix:** [specific change to make]
- **Impact:** High / Medium / Low
### [Issue 2: title]
...
## Recommendations (ranked by impact)
1. [Highest-impact change]
2. [Next change]
3. ......
Analyze recent conversations, meetings, and workflows to detect recurring patterns that should become formalized skills. Use when Lennart asks to find skill gaps, review what should be automated, identify missing skills, audit the skill library, check for patterns that need skills, or says things like "what am I doing repeatedly that we should automate" or "what skills am I missing". Also trigger proactively during weekly reviews or when the same type of request appears multiple times in a session.
Triggers: "What skills am I missing?"
Data sources: Hedy, Close, memory
Scans recent activity across Hedy meetings, Close CRM, memory files, context files, and the existing skill library to identify recurring workflows that should be turned into formalized skills. The goal is to make the EA smarter over time by catching patterns before Lennart has to ask twice.
The detector looks for three types of gaps:
GetSessions -- list recent meetings (past 2-4 weeks)GetSessionDetails -- get full transcript (transcript or cleaned_transcript). Extract topics, action items, and recurring themes from the raw conversation. Do not rely on Hedy's recap or meeting_minutes.GetHighlights -- can be used as cross-reference but transcript is primaryGetToDos -- outstanding tasks that might indicate repeated workflows (fallback: extract from transcripts)MEMORY.md and any topic-specific memory files.claude/skills/*/SKILL.md frontmatter for name and descriptioncontext/processes.md -- formalized processes that may lack automationcontext/calendar.md -- recurring cadences that could be skill-drivencontext/current-priorities.md -- priority areas where skills would have the most impactactivity_search -- recent activity types and frequencyRead all SKILL.md files in .claude/skills/ and build a coverage map:
Skill Name | Covers | Trigger Phrases
...Quarterly strategy review and strategic decision support. Use when Lennart asks for a quarterly review, strategy assessment, market positioning analysis, strategic decision, or wants to update the growth thesis. Connects KPI trends, pipeline shifts, win/loss patterns, competitive signals, and market changes into strategic recommendations.
Triggers: "Quarterly review", "strategic assessment"
Data sources: All context + data
Connects operational data with strategic context to support quarterly reviews and strategic decisions. Goes beyond the operations-health dashboard (which shows how the business is running today) to answer "are we pointed in the right direction and should we change course?"
Use this skill when Lennart says:
Always read before analysis:
context/strategy.md -- Growth thesis, market position, strategic bets, competitive moatcontext/goals.md -- Quarterly goals and targetscontext/kpis.md -- KPI targets and thresholdscontext/risks.md -- Active business riskscontext/competitors.md -- Competitive landscapecontext/pipeline.md -- Pipeline snapshotcontext/finance.md -- Cost structure and marginscontext/financial-planning.md -- Revenue model and scenarioscontext/clients.md -- Customer tiers and concentrationcontext/team-capabilities.md -- Team skills and gapscontext/current-priorities.md -- Current focus areascontext/marketing.md -- Content performance and attributionfind_opportunities -- all deals (active, won, lost) for trend analysisactivity_search -- sales activity volume and patternsaggregation -- deal metrics over timeGET /projects -- project portfolio evolutionGET /activities -- utilization trendsGET /invoices -- revenue trendssearchJiraIssuesUsingJql -- delivery velocity, issue resolution trendsapollo_mixed_companies_search -- market sizing, competitor signalsWebSearch -- industry trends, competitor news, market developmentsDelegate data gathering to parallel agents and curl calls.
Pattern:
...
"Switch between model profiles (Claude direct, OpenRouter free/paid/turbo). Use when hitting rate limits, when Lennart says switch models, use free models, use cheaper models, go faster, or go back to Claude."
Triggers: n/a
Data sources: n/a
Manages which AI model profile the EA uses.
| Profile | Main Model | Subagents | Cost | Best For |
|---|---|---|---|---|
claude | Claude direct (subscription) | Claude | Subscription | Best quality, complex EA tasks |
paid | Claude Sonnet (via OpenRouter) | Qwen 3.6 free | ~$0.003/1K tok | Good quality, Claude limits hit |
turbo | Claude Sonnet :nitro (via OpenRouter) | Qwen 3.6 free | ~$0.005/1K tok | Speed-critical work |
free | Qwen 3.6 Plus | Nemotron 120B | $0 | Emergency fallback |
:nitro — routes to fastest provider (used in turbo profile):floor — routes to cheapest provider:free — free tier modelsUse when:
/check-usage reports critical usage levelscd "g:/My Drive/Lennart/Claude Projects/1-LHE Executive Assisant (EA)" && python scripts/model_manager.py statusReports: current profile, Claude usage %, OpenRouter balance + warnings.
If Lennart specified a profile: use that directly.
If auto-selecting: run auto command which uses cached Claude usage data:
cd "g:/My Drive/Lennart/Claude Projects/1-LHE Executive Assisant (EA)" && python scripts/model_manager.py autoThe auto logic:
paid (or free if OR balance < $2)claudeManual override:
cd "g:/My Drive/Lennart/Claude Projects/1-LHE Executive Assisant (EA)" && python scripts/model_manager.py switch <profile>Where <profile> is claude, free, paid, or turbo.
Tell Lennart:
"Switched to {profile}. Restart the session to apply."
If there's urgent work in progress, offer to save context to data/session-briefing.md before restart.
...
"[DEPRECATED] Replaced by c-suite-dashboard. Use c-suite-dashboard for the weekly executive report. The CFO dashboard mode covers all financial analysis from this report. Archived copy in archives/skills/top-management-report/."
Triggers: "MD report", "cash position"
Data sources: Qonto, Close, Moco, Hedy
This skill has been replaced by the c-suite-dashboard orchestrator and the individual C-level analysis skills (cfo-analysis, cro-analysis, chro-analysis, cso-analysis, cmo-analysis).
Use instead:
c-suite-dashboardcfo-analysis (Mode 0: Weekly CFO Dashboard)cro-analysis (Mode 0: Weekly CRO Dashboard)Archive: archives/skills/top-management-report/SKILL.md
Consultant and team utilization reporting from Moco. Shows billable ratios, capacity gaps, bench risk, and resource forecast. Use when Lennart asks about utilization, billable hours, team capacity, who is available, bench, resource planning, or consultant workload.
Triggers: "Utilization", "billable hours"
Data sources: Moco
Pulls time tracking and resource data from Moco to show how the consulting team's capacity is being used. Moco is the source of truth for hours logged, project assignments, and planned effort.
Use this skill when Lennart asks about utilization, billable hours, team capacity, who is available, who is on the bench, resource utilization, consultant workload, capacity planning, or anything related to how team time is being spent.
Use this skill when Lennart says:
Triggered by: "utilization report", "team utilization", "how's utilization" Shows: Team-wide billable ratio, hours logged, top/bottom performers, trends.
Triggered by: "how busy is [person]", "what is [person] working on", "[person]'s utilization" Shows: One consultant's hours breakdown by project, billable ratio, trend, upcoming schedule.
Triggered by: "project hours", "budget burn", "hours consumed", "project utilization" Shows: Hours consumed vs. budget per project, burn rate, projected overrun/underrun.
Triggered by: "who has capacity", "who is available", "resource forecast", "staffing", "bench" Shows: Forward-looking view of planned hours vs. available capacity per consultant.
https://h2wconsultinggmbh.mocoapp.com/api/v1Authorization: Token token={MOCO_API_KEY}&page=2, etc.Report endpoints (use FIRST, pre-calculated data):
GET /report/utilization?from=YYYY-MM-DD&to=YYYY-MM-DD -- daily billable/unbillable hours per user WITH target hours. Eliminates manual ratio calculation....
Self-improvement loop for the Obsidian vault. Detects new entities in daily logs and transcripts, backfills missing Dataview inline fields on opportunities, refreshes MOCs, validates wiki-aliases, and proposes structural improvements. Use when Lennart says curate vault, vault hygiene, refresh mocs, learn from vault, vault audit, or weekly automatically via cron/closeout.
Triggers: n/a
Data sources: n/a
Purpose: make the vault measurably better every week without Lennart doing it by hand. Every run produces (1) automatic fixes and (2) a short proposals list for human review.
data/daily-log/ and data/hedy-transcripts/ for capitalized name patterns and company-suffix patterns (GmbH, AG, Inc, LLC, Ltd, KG)references/wiki-aliases.yamlaliases:: append the variant to that file's frontmatter automatically and rebuild aliases via python scripts/wiki.py build-aliasesBefore backfilling, pull current HealthTier values from Close CRM via the deal-health-scoring skill output (or the Close custom field cf_HealthTier if cached). Map by opportunity lead name. Only writes to empty HealthTier:: fields — never overwrites.
Exclusion list for entity discovery (Pass 1) — do not flag as unknown:
For every file in opportunities/:
**Stage:**, **Value:**, **Owner:**)Stage::, Value::, Owner::, Customer::, HealthTier:: inline field is missing, inject it under the H1HealthTier from Close CRM custom field if available, otherwise leave blankFor each MOC in references/mocs/:
...
"[triggers: /visualize, /visual-plan, /visual-slides, /diff-review, /plan-review, /project-recap, /fact-check, /share, visualize, diagram, visual, interactive HTML, architecture diagram, data flow, slide deck, visual plan]"
Triggers: n/a
Data sources: n/a
Transform technical content into styled, interactive HTML pages. 8 modes covering diagrams, plans, slides, diffs, reviews, recaps, and sharing.
| Command | Purpose | Template |
|---|---|---|
/visualize | Diagrams, architecture, data flows | Auto-select (see routing) |
/visual-plan | Implementation plans | timeline + css-grid-cards |
/visual-slides | Full-viewport slide decks | slide-deck |
/diff-review | Side-by-side comparisons | diff-review |
/plan-review | Audit plan vs. codebase | composite (checklist + mermaid) |
/project-recap | Context-switch mental model | composite (timeline + cards + mermaid) |
/fact-check | Verify doc claims vs. reality | data-table (claim vs. evidence) |
/share | Deploy to Cloudflare Pages | N/A |
--client <slug>: H2W brand, save to client-facing path--aesthetic <preset>: blueprint, editorial, paper, monochrome (internal only)--no-save: Skip saving, just open in browserAnalyze the content and select the best template:
| If content involves... | Use template | Technology |
|---|---|---|
| Flowcharts, pipelines, sequences, decision trees | mermaid-diagram | Mermaid.js |
| Architecture components, system overview | css-grid-cards | CSS Grid |
| ER diagrams, state machines | mermaid-diagram | Mermaid.js |
| Tabular data (4+ rows or 3+ cols) | data-table | HTML table |
| KPIs, metrics, charts | chart-dashboard | Chart.js |
| Chronological events | timeline | CSS timeline |
| Multiple content types | composite | Combined |
--client forces h2w-brand, otherwise use --aesthetic flag or default to blueprintfrom visualizations.engine import VisualEngine
engine = VisualEngine(
title="Page Title",
subtitle="Optional subtitle or date",
aesthetic="blueprint", # or editorial, paper, monochrome, h2w-brand
)
engine.render(
template="mermaid-diagram", # template name without .html
content={
"mermaid_code": "flowchart TD\n A-->B",
"sections": "", # additional HTML sections below the main content
},
...Generate a comprehensive weekly review report for Friday leadership meetings. Pulls from Hedy, Close CRM, Jira, and Slack to synthesize the week into an actionable briefing. Use when Lennart says weekly review, Friday briefing, what happened this week, weekly report, or week in review.
Triggers: "Weekly review", "Friday briefing"
Data sources: Hedy, Close, Jira, Slack
Comprehensive weekly review report for Friday morning leadership meetings. Pulls from all connected data sources (Hedy meetings, Close CRM, Jira, Slack) and synthesizes the week's activity into a single actionable briefing.
This skill connects the dots across meetings, sales, delivery, and team activity to surface what matters most and what needs attention going into the next week.
Use this skill when Lennart says:
All tools are connected via MCP through Claude.ai. No API keys needed.
GetSessions -- list meetings in a date rangeGetSessionDetails -- full transcript, recap, highlights, action items per meetingGetToDos -- all outstanding to-dos across sessionsGetSessionHighlights -- highlights from a specific sessionGetSessionToDos -- to-dos from a specific sessionfind_opportunities -- all active opportunitiesactivity_search -- emails, calls, tasks, meetings, notes (filter by date)aggregation -- activity counts by type, user, time bucketfind_pipelines_and_opportunity_statuses -- pipeline stagesorg_users -- map user IDs to namessearchJiraIssuesUsingJql -- issues updated/created/resolved this weekgetJiraIssue -- issue detailsgetVisibleJiraProjects -- active projectsslack_search_public_and_private -- search for key topics discussed this weekslack_read_channel -- read recent messages from key channelsDelegate all data gathering to subagents launched in parallel. The main conversation handles synthesis, analysis, cross-referencing, and report generation.
Launch three research tasks simultaneously:
Agent 1: Meetings (Hedy) Use Hedy tools directly (not the research-agent, since it does not have Hedy tool access).
GetSessions with limit 20 to get all meetings from the past 7 days...
Capture weekly business metrics snapshot for trend analysis. Run after Monday reports to save key numbers (cash, pipeline, utilization, delivery, Labs) to reports/internal/snapshots/. Use when Lennart says save the snapshot, capture this week's numbers, weekly snapshot, save metrics, or after generating Monday reports.
Triggers: "Save the snapshot", "capture this week's numbers"
Data sources: Qonto, Close, Moco, Jira
Captures a structured snapshot of key business metrics and saves it to reports/internal/snapshots/YYYY-MM-DD.md. This creates a time series that all other reports can reference for trend analysis, week-over-week comparisons, and quarterly reviews.
Run this after the Monday reports (top-management-report + monday-briefing) to persist the numbers that were pulled. Without snapshots, each report starts from zero with no historical context.
Use this skill when Lennart says:
Also trigger automatically when both Monday reports have been generated in the same session.
Read before pulling data:
context/kpis.md -- KPI definitions and thresholds (GREEN/YELLOW/RED)context/finance.md -- Cost structure baselinecontext/financial-planning.md -- Cash flow thresholdsCheck for the most recent snapshot:
reports/internal/snapshots/ -- Read the latest file to compare week-over-weekGET /organization -- current balances for both entitiesAuthorization: $QONTO_H2W_AUTHAuthorization: $QONTO_BLC_AUTHfind_opportunities -- all active deals with values and stagesfind_pipelines_and_opportunity_statuses -- stage definitionsaggregation -- activity counts for the weekGET /report/utilization?from={monday}&to={friday} -- pre-calculated billable/unbillable per user WITH target hours (use instead of raw activities for utilization metrics)GET /activities?from={monday}&to={friday} -- raw hours (use for project-level breakdown if needed)GET /projects?include_archived=false -- active project count and budgetsGET /users -- team member countGET /report/absences?year=YYYY&active=true -- vacation/sick data for capacity contextsearchJiraIssuesUsingJql -- issues created, resolved, blockersLaunch 3 parallel data gathering streams:
Stream 1: Financial
...
Workforce planning, team capacity analysis, and people decisions. Use when Lennart asks about staffing, hiring, team redeployment, skill gaps, bench risk, Neudorff impact on team, who to assign to new projects, or workforce scenarios. Goes beyond utilization-report (which shows current hours) to answer forward-looking people questions.
Triggers: "Staffing", "[[w-neudorff
Data sources: Moco, Close
Acts as a virtual Head of People Operations for Lennart. Goes beyond the utilization-report (which shows current-state hours) to provide forward-looking workforce analysis, redeployment planning, and hiring/reduction decisions.
Use this skill when Lennart says:
Always read before analysis:
context/team-capabilities.md -- Skills matrix, capacity profile, exposure analysis, key-person dependencies. Includes Historical Utilization Baselines section with per-person annual hours, external %, customer concentration, and vacation/sick patterns from Octaved (Oct 2022 - Feb 2026). Use these instead of generic assumptions.context/team.md -- Team structure and key peoplecontext/team-assignments.md -- Current project assignments (from Moco)context/finance.md -- Internal rates, cost structure, actual rate realization, internal cost ratio trendcontext/kpis.md -- Utilization targets and thresholdscontext/current-priorities.md -- Strategic contexthttps://h2wconsultinggmbh.mocoapp.com/api/v1Authorization: Token token={MOCO_API_KEY}Report endpoints (use FIRST, pre-calculated data):
GET /report/utilization?from=&to= -- daily billable/unbillable per user WITH target hoursGET /report/planned_vs_tracked?from=&to= -- planned vs actual hours by user+project with delta and quotaGET /report/absences?year=YYYY&active=true -- per-user vacation (total/used/planned) + sick daysGET /users/{id}/performance_report?year=YYYY -- annual+monthly: target, tracked, billable, variation per userGET /projects/{id}/report -- budget, hours, costs, profitability per projectRaw data endpoints:
GET /users -- team members (id, name, unit, active)...
Prepare agendas, materials, checklists, and follow-up documents for customer workshops. Use when Lennart asks to prepare a workshop, draft a workshop agenda, write up workshop results, create a workshop proposal, or build a prep checklist for any AI discovery, ERP, or transformation workshop.
Triggers: "Prepare workshop for [company]"
Data sources: Context files
Prepare agendas, materials, and follow-up documents for customer workshops.
Use this skill when Lennart says:
Adapt per customer and workshop type, but follow this flow:
1. Context and objectives (15-30 min)
- Why we are here, what we aim to achieve
- Confirm scope and expected outcomes
2. Current situation and challenges (30-60 min)
- System landscape overview
- Current processes and pain points
- What has been tried, what worked, what did not
3. Process walkthrough (60-120 min)
- Walk through relevant processes in detail
- Identify gaps, manual steps, workarounds
- Capture dependencies and integration points
4. Identification of opportunities (30-60 min)
...Gathers pipeline, deal, contact, and activity data from Close CRM. Use for pipeline snapshots, deal research, sales activity metrics, and lead intelligence. Returns structured raw data without analysis.
17 tools: Read, Grep, Glob, mcp__claude_ai_Close__search ...
Gathers financial data from Qonto bank API and Moco invoices. Use for cash position, transaction analysis, invoice tracking, and receivables. Returns structured raw data without analysis.
4 tools: Read, Grep, Glob, Bash
Gathers meeting data from Hedy. Use for meeting transcripts, action items, decisions, highlights, and session summaries. Returns structured raw data without analysis.
13 tools: Read, Grep, Glob, mcp__claude_ai_Hedy__GetSessions ...
Gathers project delivery data from Jira. Use for issue tracking, delivery status, sprint progress, blockers, and workload analysis. Returns structured raw data without analysis.
9 tools: Read, Grep, Glob, mcp__claude_ai_Atlassian__searchJiraIssuesUsingJql ...
Gathers email, calendar, contacts, and task data from Microsoft 365 via Graph REST API. Use for inbox scanning, calendar lookups, email search, and task retrieval. Returns structured raw data without analysis.
4 tools: Read, Grep, Glob, Bash
Gathers project, resource, and time tracking data from Moco API. Use for utilization analysis, project budgets, team assignments, and resource planning. Returns structured raw data without analysis.
4 tools: Read, Grep, Glob, Bash
Gathers data from Close CRM, Jira, Apollo.io, and web sources. Use when researching companies, deals, pipeline data, or market intelligence. Returns structured raw findings without analysis or recommendations.
34 tools: Read, Grep, Glob, WebSearch ...
- Our clients are industrial manufacturing companies
.claude/rules/business-context.md- Lead with the answer or recommendation
.claude/rules/communication-style.md- Never send emails without Lennart's explicit approval. Always create as draft.
.claude/rules/email-drafting.mdAll PDF reports must follow `references/report-design-standards.md`. Read that file before generating any PDF.
.claude/rules/report-generation.md- Always structure proposals with clear scope, deliverables, timeline, and commercial terms
.claude/rules/sales-and-proposals.mdThe vault is an Obsidian-browsable knowledge graph. Its usefulness depends on entity files being linked to each other, so backlinks and graph view reveal relationships. These rules govern how to reference entities when writing or editing files.
.claude/rules/wiki-linking.mdm365.shm365_classify.pym365_extract_enrichment.pym365_fetch_body.pym365_history.pym365_index.pym365_summarize.pymail_format.pycreate_tsurumi_ai_order_intake_moco.pyparse_moco_utilization.pylinkedin_client_map.pylinkedin_dormant.pylinkedin_infor_map.pylinkedin_warm_intros.pybackfill_opp_custom_fields.pycreate_crm_smart_views.pycreate_deal_health_smart_views.pydeal_health_score.pydeal_health_score_config.pygemini.shmodel_manager.pyvault_curator_activity_rollout.pyvault_curator_backfill.pywiki.pybrand_image.pybuild_jbc_proposal.pybuild_tsurumi_ai_order_intake_proposal.pygenerate_pptx.pymd2pdf.pyscraper.pyattach_jbc_pdf.shcheck_staleness.pycheck_write_protected.pycontext_refresh.pycreate_ermco_activity.pygen_ea_architecture.pyq1_scan.pyPipeline, leads, opportunities, activities, contacts, smart views.
Project tracking, sprint data, delivery status, knowledge base.
Company and contact enrichment, prospecting, email discovery.
Meeting transcripts, session highlights, to-dos, topics.
Channel history, DMs, search across workspace.
Up-to-date library documentation lookup.
D1, R2, KV, Workers, Hyperdrive management.
Job posting and resume data.
Weekly executive briefing for Monday morning BD-PMO call.
Combined CFO/CRO/CHRO/CSO/CMO weekly dashboard.
Deep stakeholder research on active deals.
Status reports sent to clients.
Deal-stage positioning decks for enterprise pursuits.
Formal proposals and statements of work.
Internal + customer-facing account review materials.
| Date | Tag | Decision |
|---|---|---|
| 2026-04-08 | TOOLING | Obsidian wiki-linking Phase 2: scripts/wiki.py CLI (lint-schema / build-aliases / dedupe-people / gap-report / link-backfill), references/wiki-schema.md (entity types + frontmatter), .claude/rules/wiki-linking.md (first-mention wikilinks in wiki layer, relative markdown in config layer). Obsidian config flipped to useMarkdownLinks: false, newLinkFormat: "shortest". Applied lint-schema defaults to 113 files, manually enriched aliases for 17 top entities, built 211-entry alias registry, backfilled 1193 first-mention wikilinks across 167 files. Inbound link density now dense around trojan-technologies (76), w-neudorff (74), tsurumi (71), knowkit (63), meisterwerke-schulte (44). Two bugs caught and fixed during apply: every-mention-linking and self-linking. Commit eb8269e. |
| 2026-04-08 | TOOLING | Untrack Obsidian runtime state (.obsidian/workspace.json, graph.json) and raw Hedy API dumps (data/hedy-transcripts/*.json) — these produce constant phantom diffs and pollute history. Processed .md files stay tracked. Commit 531567b. |
| 2026-04-08 | TOOLING | EA project showcase site v2 at webapp/showcase/. Self-regenerating single-file HTML (1036 KB, password ea-showcase2026) parsing 81 skills + daily logs + decisions + report PDFs + git log + memory + context files + 5-layer architecture diagram + clickable skill modal with body drill-down + daily drivers strip. 17 sections. Canonical pattern matches client-website skill (Netlify, base64 assets, SHA-256 gate). Target URL ea.blconsultingservice.com. New project-showcase skill wires regenerate/deploy triggers. Commits d6325d7 (v1), 13e150c (skill). Deploy still manual (Netlify init + custom domain). |
| 2026-04-08 | TOOLING | context/index.md is the single source of truth for context files. YAML frontmatter (auto_refresh, refresh_interval_days, source) drives staleness hook, write-protection hook, and context-refresh skill. CLAUDE.md slimmed 191→81 lines. Adding/moving/renaming a context file now requires touching only the index. |
| 2026-04-07 | STRATEGY | Q2 2026 goals locked: (1) replace [[w-neudorff |
| 2026-04-07 | FINANCE | Datev is the source of truth for invoicing and AR aging, NOT Moco. Moco invoice list is downstream and not authoritative. Future scans must pull AR from Datev. Saved as permanent memory. |
| 2026-04-07 | TOOLING | Deal Health Scoring system: rule-based stage-aware scoring for ~90 qualified-funnel opps (Interested but need time through Verge of closing). 4 Close custom fields, 3 Smart Views, wired into monday-sequence + deal-pulse + cro-analysis. Lead scoring for early stages (Diverse conversation, Potential for Discussion) deferred. |
| 2026-04-06 | TOOLING | Session-start timezone check: use PowerShell TimeZoneInfo (not bash TZ=), shows ET always + detects travel. Saved to memory. |
| 2026-04-02 | TOOLING | Relationship intelligence system: 41-contact registry, role-based cadence defaults, direct+team touch tracking, standalone skill + daily-scan integration. Pure file-based (data/relationships/*.md). |
| 2026-04-02 | SALES | [[trojan-technologies |
| 2026-04-02 | OPS | 20% PM overhead standard on all project quotes. Applied retroactively to Trojan Dev Training (USD 15,600 to USD 18,720). |
| 2026-04-02 | TOOLING | Close Reporting API integrated into 3 skills (cro-analysis, sales-intelligence, c-suite-dashboard). Native funnel, status transition, and activity metrics replace manual computation. Org/pipeline IDs hardcoded. |
Standing meetings and recurring events
Communication channels across systems
Customer list and relationship status
Per-contact communication preferences
Competitive landscape
Contract templates and standing agreements
What Lennart is focused on right now
Recurring deal structures and pricing
Cash position, budgets, P&L structure
Forecast and scenario models
Quarterly goals and milestones
Ideal customer profile
Index of all context files with frontmatter metadata
Tracked metrics and thresholds
Marketing channels and activity
Context file
Service catalog and packages
Partner ecosystem (Infor, AWS, etc.)
Active deals and stage
Internal operating processes
Who Lennart is, entities, revenue streams, tools
Active delivery projects
Interaction history with key contacts
Planned revenue ramps and cliffs
Risk register
Sales process and qualification criteria
Target SEO terms for content
Key external decision-makers
Strategic bets and positioning
Who is staffed on which project
Skills matrix across consultants
Team changes and tenure
Current team roster
Context file
Why deals were won or lost
Current work in flight
Memory persists across Claude Code sessions. Stored in the per-project memory directory outside the repo.
project_labs_poc_roles.mdfeedback_scalable_structures.mdfeedback_superpowers_proactive.mdfeedback_daily_log_mandatory.mdfeedback_no_close_field_enrichment.mdfeedback_moco_deal_category.mdfeedback_moco_company_search_all_types.mdfeedback_moco_company_type.mdproject_evaluation_goals_rollout.mdproject_lucas_departed.mdproject_web_ea.mdproject_pptx_generator.mdreference_cloudflare_account.mdreference_netlify_custom_domain_swap.mdfeedback_client_website_quality.mdproject_client_websites.mdproject_dns_automation.mdreference_hedy_api.mdreference_hedy_configuration.mdfeedback_hedy_indexing_trust_calendar.mdfeedback_dual_machine_portability.mdfeedback_weekly_insights.mdfeedback_close_custom_fields_prefix.mdfeedback_meeting_workflow_improvements.mdfeedback_disabled_skill_read_directly.mdfeedback_business_partner_question_framing.mdfeedback_time_aware_greetings.mdfeedback_german_text_quality.mdEvery Monday morning, Lennart runs a single command: "Monday sequence". What happens next is a chain of 20+ skills executing in rough priority order, producing one combined executive brief he reads over coffee before the 7 AM BD-PMO call.
context-refresh pulls fresh data from Moco, Close, and Qonto into local snapshot files.weekly-snapshot captures this week's numbers as a timestamped artifact so trends stay comparable across weeks.deal-health-scoring re-scores all 90+ qualified opportunities on a 0-100 scale, tiers them Green/Yellow/Red/Stalled, and updates Close smart views.deal-pulse watches the last 7 days of email for deal signals and flags overdue follow-ups.relationship-pulse checks interaction frequency with key contacts and raises alerts for relationships going cold.alert-check does a quick scan of local files for red flags (stale drafts, missing MSAs, overdue tasks).c-suite-dashboard chains five C-level analysis skills in parallel:cfo-analysis — cash position, margin, scenario modelingcro-analysis — pipeline health, revenue forecast, deal prioritizationchro-analysis — team health, retention risks, bench costcso-analysis — strategic bets, competitive intel, risk registercmo-analysis — marketing performance, content output, brand healthmonday-briefing stitches it all together into one structured report with the top 3-5 actions for the week.inbox-drafts triages the inbox and drafts replies to anything needing a response.One PDF open on the left monitor, an inbox of pre-drafted replies on the right. He spends 10 minutes reviewing the brief, approves or edits the drafts, and walks into the 7 AM BD-PMO call with a complete picture of pipeline, delivery, cash, people, and content — across three entities.
Without this sequence, the same prep would take 2-3 hours of pulling reports from Close, Moco, Qonto, Jira, and email. And it would happen inconsistently, because nobody wants to do 2-3 hours of manual report-stitching every Monday at 6 AM.
The Monday sequence is the canonical example of why formalizing workflows as skills matters: the effort to build it once is large, but the weekly payoff is compounding.
A sales deal in the EA project moves through six formalized stages, each with its own skill. The EA handles the data work, the research, the drafting, and the tracking. Lennart handles the conversations and the decisions.
prospecting-engine builds target account lists from Apollo.io filtered against the ICP in context/icp.md. It enriches with contact data and stages companies for outreach in Close CRM.meeting-briefing-notes researches any prospect before a first conversation — Apollo enrichment, web scraping, historical CRM activity, past meeting transcripts if any.sales-discovery-debrief runs after every discovery call. It reads the Hedy transcript, extracts pain points, budget signals, timeline, decision-makers, and writes back to Close CRM. It drafts the follow-up email in the right language and tone.customer-intelligence builds a deep stakeholder map and profile by pulling from Close, Jira, Moco, Hedy, and Apollo.proposal-sow-generation drafts a full proposal or SOW using templates from references/proposal-creation-guide.md. It reads the deal context from Close, the existing scope notes, and the ICP file, and produces a structured document with commercial terms.client-website (optional) generates a password-protected microsite at {client-slug}.blconsultingservice.com with the positioning story, references, and an inline CTA to the Moco customer approval link.deal-pulse monitors email and calendar for replies. If a deal goes silent, it surfaces the need for a follow-up.sales-outreach drafts re-engagement emails, follow-ups, or check-in messages tailored to deal stage and recipient relationship.deal-health-scoring tiers every qualified deal. Red and Stalled deals get flagged in the Monday briefing.h2w-labs-playbook runs the H2W Labs conversion playbook for innovation/AI deals — opportunity brief, demo prep, proposal, follow-up cadence.moco-crm-sync links the closed Close opportunity to a new Moco project.labs-demo-delivery or delivery skills take over for execution.qbr-client-review runs quarterly for existing customers.Every skill in the chain reads from and writes to the same source of truth. Close is the system of record for pipeline. Moco is the system of record for active project tracking. Hedy is the source of truth for meeting transcripts. The EA never duplicates data — it enriches the authoritative system and moves on.
The result: a deal's full history lives in Close, not in a pile of scattered emails, documents, and notes. And when Lennart asks "where are we with Trojan?" the EA can answer in seconds by reading Close, cross-referencing Jira for delivery status, and checking Moco for any active project work.
The EA runs a complete marketing content operation — from topic generation through drafting, review, scheduling, analytics, and newsletter assembly. Every piece of published content flows through the same pipeline.
marketing-content-pipeline turns project experiences, customer conversations, and operational insights into content topics. It mines Hedy transcripts, daily logs, and decision logs for stories worth telling — always anonymized, always with business-value framing.content-engine is the orchestrator. It checks queue health, drafts in batch, and maintains a rolling 3-day approved buffer so publishing never runs dry.blog-writer produces SEO-targeted blog articles in German and English with meta descriptions and keyword targeting.conversion-copywriting sharpens existing copy, tightens CTAs, improves hook strength.copy-editing runs a seven-sweep quality check (voice, tone, brand, grammar, umlauts, formatting, fact-check).content-review presents drafted posts in batch for Lennart's approval. It tracks edits for voice calibration — so the next batch drafts closer to how he actually writes.data/content-queue/approved/. A lightweight LinkedIn publishing workflow picks up posts by date.content-analytics tracks what's working — engagement per post, which topics land, weekly and monthly reports with pipeline attribution.cmo-analysis rolls this into the weekly CMO dashboard so content ROI stays visible at the executive level.newsletter-assembler once a month curates the best published pieces, assembles a bilingual newsletter, and creates the Close CRM email templates and workflows for sending.The trickiest part of content generation isn't writing — it's sounding like Lennart. The EA handles this in three ways:
.claude/rules/communication-style.md enforce the hard constraints (no emojis, no em dashes, no buzzwords, Eastern Time for all times, proper umlauts).context/profile.md, context/strategy.md, context/icp.md) keep the EA aware of positioning and audience.content-review captures what Lennart changed in past drafts and feeds it back into the next generation cycle.Over time, the gap between first draft and publish narrows. That's the feedback loop working.
A typical manufacturing consulting firm has either (a) a full-time marketing hire or (b) zero content output. The EA lets a single operator run a credible two-track content program (LinkedIn daily, blog weekly, newsletter monthly) as a side effort to actual client work.
A few principles hold this whole project together. They aren't written in the rules files because they're more fundamental than that — they shape why the project exists the way it does.
If you do something twice, write it down. If you do it three times, make it a skill. Every skill in this project started as a manual workflow Lennart ran often enough to be annoyed by. Formalization is the unit of automation.
Close is the system of record for pipeline. Moco for projects. Datev for invoices. Jira for delivery. Hedy for meeting transcripts. The EA reads from and writes to these systems directly. It does not build parallel databases. It does not cache state that goes stale. If the data lives somewhere authoritative, that's where it stays.
Emails, Close updates, Moco entries — the EA creates drafts. Lennart approves. This is a hard rule, not a suggestion. It's the difference between an assistant that speeds you up and an assistant you can't trust.
Most of what makes the EA useful is not the skills or the prompts. It's the 36+ context files describing who Lennart is, what the business is, who the clients are, who the team is, what the goals are, what's currently on fire. A clever prompt against a thin context produces generic output. A simple prompt against a rich context produces output that reads like it came from someone who actually knows the business.
Everything is markdown. Files in a git repository. Wiki-linked for Obsidian browsing. Grep-able, diff-able, versionable. When you look at this project you can see every line of every rule and every skill. No black boxes.
Claude Code has persistent memory across conversations. But memory is not a log — it's a curated set of facts about Lennart, the project, and the work. Saving too much makes it noisy. Saving too little means the assistant re-learns the same things every session. The memory/ directory is a hand-curated index.
The EA is fast at generation. That's the easy part. The actual constraint is Lennart's attention for review. So the skills that matter most are the ones that collapse many small decisions into a few big ones — the Monday brief, the inbox drafts, the content review batch. Each one trades dozens of small context switches for one focused review session.
The EA is not autonomous. It does not close deals, run meetings, or make strategic decisions. It gathers data, drafts artifacts, and flags things that need attention. That's the right scope. The moment a tool claims to do more than it actually does, trust in it collapses — and the project grinds to a halt.
The goal isn't to replace judgment. The goal is to load as little friction as possible around the parts of the work that require judgment.