retrolisted

# retro — team-coached retrospective The continuous-improvement ritual. It does NOT make the model smarter; it improves the **context** the team runs with next time (skills, guardrails, role briefs, routing) — and only via durable artifacts that get **retrieved** at the right moment. Canonical mechanics: wiki `experience/methodology/team-coached-retrospective-loop`. ## The outcome-log (the signal — feed it continuously) The loop is worthless without a signal. During normal work, append one line to the project's outcome-log (`<repo>/.memory/outcome-log.md`, or the wiki's `Wiki/journal/outcome-log.md` for wiki work) whenever something is worth learning: ``` <date> · <type> · <one line: what happened> → <delta or "none"> ``` Entry **types**: `outcome` (verify green/red), `review-catch` (a reviewer/peer caught an error before it landed), `provenance`/`redaction` (a sourcing/secret slip), **`user-correction`** (a behavior the user criticized — verbatim gist). User corrections are first-class: capture them the moment they happen, not just at /retro. ## The ritual (when /retro is invoked) 1. **Read the signal** — the recent outcome-log entries + the relevant diffs/commits. 2. **Enact role-coach** (advisory; tools: Read/Grep/Glob). For each recurring or high-cost entry, propose a **concrete delta** — never "be more careful": - a process **gate** (a check added to a skill/workflow), - a **skill edit** (hand to skill-smith), - a **guardrail** (a rule that prevents th