ilk-feedbacklisted

# ilk-feedback — postmortem skill for ilk-loop runs A read-only triage skill that turns the structured logs already produced by `run_ilk_loop_claude.ps1` into: 1. A **classification** of how the run ended (one of 12 taxonomy labels). 2. **Parameter recommendations** for the next launch (consumed by `ilk-launcher` Step 1.5). 3. A **markdown report** persisted to disk for trend analysis. 4. A **chat summary** with a 3-way choice — (a) resume / (b) bump-then-resume / (c) fix code first. ## When to use - The user asks for a postmortem, debrief, or "what went wrong" about a recent ilk run. Common phrasings (English + 中文): - English: `postmortem`, `postmortem on <project>`, `debrief`, `debrief the last run`, `ilk debrief`, `do a postmortem`, `why did ilk stop`, `why did <project> stop`, `what went wrong`, `what went wrong with ilk`, `what happened on the last run`, `feedback on the last run`, `/ilk-feedback` - 中文：`复盘`、`复盘一下`、`复盘 ilk`、`ilk 反馈`、 `为什么 ilk 中断`、`ilk 怎么停了` - Codex: "postmortem the last run", "why did the loop stop", "ilk feedback" - A ilk window just exited (clean or otherwise) and the user wants feedback before re-launching. - The user wants the agent to consider history before launching ilk again — `ilk-launcher` Step 1.5 invokes this implicitly by reading postmortem frontmatter. ## Architecture `<skill-root>` below means the installed skills base directory — `~/.cursor/skills/` (Cursor), `~/.claude/skills/` (Claude Cod