Manage ~/.fdestack/learnings.jsonl — the cross-customer pattern library. List, search, prune stale entries, edit incorrect insights, or export the file for sharing with another FDE. No write to customer files; this is meta over the learnings file itself. Proactively invoke this skill when the FDE asks "what have we learned", "what patterns have we seen", "didn't we hit this before", or wants to share/export accumulated insights. Distinct from /customer-context Step 3, which only surfaces the top 3 most recent — /learnings is the full management UI. Usage: /learnings [list | search <kw> | prune | edit | export] Voice triggers: "what have we learned", "show learnings", "prune learnings", "have we seen this before".
Process any inbound source of customer context — meeting transcript, prior FDE's handoff notes, AE/sales briefing, customer-side RFP response, prior Discovery Workshop output, or your own raw notes. Extracts new stack info, stakeholders, stated problem, real problem, open questions, and context conflicts. Updates customer context files and commits. Proactively invoke this skill (do NOT just summarize) when the user pastes call notes, meeting transcripts, AE briefings, RFP responses, handoff docs, or any inbound document about a customer. Triggers on phrases like "I just had a meeting", "here are notes from", "process this call", or any content that looks like raw or synthesized customer information. Usage: /discovery <customer-name> <paste any source doc> Voice triggers: "process the call", "I just had a meeting", "process these notes", "discover for <customer>".
Force the wedge question: smallest thing buildable this week that proves the core value. Require binary success criteria. Push vague answers to unknowns.md as HIGH so they surface in every future Session Brief. Proactively invoke this skill after /discovery when the FDE describes a problem they want to solve for the customer, references the wedge or first thing to build, or asks "what should we build first" / "what's the smallest thing". Always run before /poc or /value-frame. Usage: /scope <customer-name> <problem statement> Voice triggers: "scope this out", "what's the wedge", "smallest thing that proves value", "what should we build first".
Weekly retrospective for a single customer engagement. Reads commit history, recent discovery/scope/value-frame/poc/integrate files, and current blockers/unknowns. Produces a structured retro: what shipped, what's stuck, what we learned, what's next. Optional cross-customer learnings write-back for patterns worth remembering. Distinct from gstack's /retro (which is engineering-retro per-codebase). Proactively invoke this skill on Fridays or end-of-week, or when the FDE asks "where did the week go", "weekly retro", "what shipped this week", or seems to be wrapping up an engagement phase. Always per-customer — use /triage for the portfolio view. Usage: /engagement-retro <customer-name> [--since <YYYY-MM-DD>] Voice triggers: "weekly retro", "where did the week go", "what shipped this week", "wrap up <customer>".
Production-grade implementation. Reads scope.md + stack.md + decisions.md + value-frame.md. EXPLICITLY DOES NOT READ customers/<name>/poc/ — that directory is throwaway. POC discoveries reach /integrate via the learnings write-back that /poc performs into stack.md and decisions.md. Proactively invoke this skill after /poc has run AND its learnings write-back is complete when the FDE says "let's productionize", "time to integrate", "build the real thing", or "cleanroom rebuild". Refuse to run if no /poc exists yet — surface the gap rather than building blind. Usage: /integrate <customer-name> Voice triggers: "productionize", "build the real thing", "integrate this", "cleanroom rebuild".
Build a throwaway proof-of-concept. Speed over quality. Hardcoded values OK, ugly code OK, no error handling beyond crash prevention. This code is NEVER going to production — /integrate will rebuild fresh. The skill's most important step is the learnings write-back at the end, which routes discoveries to stack.md / decisions.md / learnings.jsonl so /integrate can use them without ever reading POC code. Proactively invoke this skill after /scope when the FDE says "let me build a quick thing", "let me try this", "time to prototype", "let's hack something together", or otherwise indicates they're about to write throwaway code for a customer. Push back if no /scope exists yet. Usage: /poc <customer-name> [--purpose value|showcase|feasibility] Voice triggers: "build a poc", "throwaway prototype", "let me try this", "hack something together".
Cross-customer dashboard. Scan every customers/<name>/ directory, surface the highest-urgency blockers and unresolved unknowns across the FDE's portfolio. Optional filter to one cohort. No session marker required — this skill is the "where do I look first today?" view. Proactively invoke this skill at the start of a session in an FDE ops repo when no specific customer has been named, OR when the FDE asks "where should I look first", "what's most urgent", "what's on my plate today", or any variant of "what should I be working on?". Usage: /triage [active|blocked|all] Voice triggers: "what's urgent", "where do I look first", "what should I work on", "what's on my plate".
Translate a scoped process inefficiency into a defensible dollar value with stated assumptions and sensitivity. Forces lever taxonomy, explicit unit economics, and a defensibility check. Vague items get pushed to unknowns.md HIGH so they surface in every future Session Brief until resolved. Proactively invoke this skill after /scope when the FDE asks "is this worth doing", mentions ROI, asks about business case, needs to prepare for an executive conversation about value, or wants to put a dollar number on the work. Especially relevant before executive readouts or renewal calls. Usage: /value-frame <customer-name> <opportunity descriptor> Voice triggers: "frame the value", "what's this worth", "build a business case", "put a number on it".