seed-workspace

Day-1 onboarding interview that maps the user's existing tools and produces Setup-Decisions.md — the per-ability decision matrix that drives everything else in the seed. Use when the user says 'discover my stack', 'set up the workspace', 'onboard me', 'what should I install', 'help me configure this seed', or whenever Setup-Decisions.md is empty / missing. Non-blocking: produces a draft the user can edit; never installs anything itself.

Sister to `harvest-learnings`. Take a staged doc from `Strategy/to integrate/`, propose specific edits to canonical strategy docs (paths read from `.skill-config.yml`), get the user's explicit per-edit approval, apply approved edits, then archive the source to `Strategy/z-archived/`. Use when the user says: 'integrate the [topic] learnings', 'integrate from to-integrate', 'apply the staged learnings', 'merge into canonical', 'fold the [client] insights into strategy', or 'time to integrate'. **NEVER writes to canonical without explicit per-edit approval.** Supports preview-only mode + partial integration. Manual invocation only.

Orchestrate deep research across meeting transcripts in the local transcript vault. Use when the user asks to analyze client meetings, review sales calls, research a topic across meetings, build client personas, export transcripts for RAG, or says 'meetings', 'transcripts', 'call review', 'client analysis'.

Per-ability decision protocol for the seed workspace. Asks 'what's the marginal gain over what your live system already does?' for any proposed local subsystem (transcript vault, CRM index, drive index, custom MCP, scripted automation). Use when the user says 'should I build X', 'do I need a local index for Y', 'what's the gain', 'is this worth setting up', or whenever a row in Setup-Decisions.md is marked needs-gain-analysis. Outputs a recommendation with reasoning, not a decision — the user decides.

Extract transferable learnings from a client engagement (or strategic prep session) and stage them in `Strategy/to integrate/` for later canonical integration. Reads engagement artifacts from `clients/<client>/`, cross-references against the canonical strategy docs (paths read from `.skill-config.yml`), and drafts a properly-sourced staging doc using the to-integrate TEMPLATE.md. Use when the user says 'harvest learnings from [engagement]', 'what did we learn from [client]', 'feedback loop time', 'lessons from [client]', 'wrap up [engagement]', or 'extract insights from this work'. Does NOT modify canonical docs — that's the sister `integrate` skill's job. Manual invocation only.

Install one operational cadence at a time in Container/cadences/. The discipline is one-at-a-time, never bundled — past four sticky cadences you're over-instrumented. Use when the user says 'install a cadence', 'add a weekly review', 'set up the keystone review', 'scaffold a Friday review', 'add a monthly retro', or any 'install one operational ritual' request. Drafts the cadence file, names the calendar slot, defines drift signals, and asks the user to commit to a 4-week trial.