co-promote

name

co-promote

description

Layer 3 promote-to-SKILL for claude-codex-orchestration. Take a candidate identified by /co-review (Durability + Impact + Scope ≥ 6), distill from descriptive to prescriptive, write the rule into SKILL.md or appropriate reference file, remove source .error-log.jsonl entries, trigger Layer 0 auto-score and README sync. Ratchet rule enforces next same-dimension session must improve or revert. Delegates to claude-codex-orchestration references/self-correction.md Layer 3.

/co-promote — Write Candidate Rule into Skill (Layer 3)

Full protocol: ~/.claude/skills/claude-codex-orchestration/references/self-correction.md §Layer 3.

Quick run

1. Take the candidate from /co-review output (must have score ≥ 6)
2. Distill from descriptive → prescriptive:
   • ❌ "Codex kept asking about scope"
   • ✅ "If Codex task spec exceeds 200 words, compress before sending"
3. Pick target file (SKILL.md for core rules; references/*.md for detailed mechanics)
4. Apply simplicity criterion: shorter rewrite at same score wins
5. Write the change; trigger Layer 0 auto-score + README sync per rule
6. Remove source .error-log.jsonl entries
7. If a Layer 2.5 .issue-log.jsonl issue matches → auto-close with comment "Addressed in <commit>. See SKILL.md §<section>."
8. Ratchet: next session scoring the same dimension must improve, else revert