codex-pr-distilllisted

# Codex PR Distill > **Requires:** the `pr-contribution-excellence` skill at `~/.claude/skills/pr-contribution-excellence/SKILL.md` (bundled with the `kookr-toolkit` plugin since 0.5 — see `docs/hooks-setup.md`). If absent, the plugin is not installed; stop and report the missing dependency rather than fabricating distilled patterns. ## When to Use This skill is invoked when `codex-pr-threshold` returns `DISTILL`. It reads all accumulated raw learnings, identifies patterns, and writes them to the appropriate output files. ## Non-Negotiable Rules | # | Rule | Violation Example | Correct Pattern | |---|------|-------------------|-----------------| | 1 | Read ALL raw learnings before distilling | Only reading the latest batch | Read entire learnings-raw.md | | 2 | Preserve prior distilled content as foundation | Overwriting previous distillation from scratch | Read existing distilled files, merge new patterns | | 3 | Separate codex-specific from general patterns | Putting "use cargo fmt" in general skill | Codex-specific → repo/openai-codex.md; General → user skill SKILL.md | | 4 | Reset learnings-raw.md after distillation | Leaving raw file to grow unboundedly | Reset to header template (6 lines) | | 5 | Update distillation_count in state.json | Forgetting to track distillation | Increment count, set last_distillation_at | | 6 | Patterns must be evidence-based | "I think PRs should..." | "PRs that {observed pattern} were {merged/rejected} — seen in #{n1}, #{n2}" | ## Dist