Codex Commit Review

Purpose

Debate committed code quality after committing, before pushing. No code modified -- report + suggest only.

When to Use

After committing code (before push). Modes: staged (pre-commit preview) or last (already-committed).

Prerequisites

• Staged: staged changes available (git diff --cached).
• Last: recent commits exist.

Runner

RUNNER="{{RUNNER_PATH}}" SKILLS_DIR="{{SKILLS_DIR}}" json_esc() { printf '%s' "$1" | node -e 'let d="";process.stdin.on("data",c=>d+=c);process.stdin.on("end",()=>process.stdout.write(JSON.stringify(d)))'; }

Critical Rules (DO NOT skip)

• Stdin: printf '%s' "$PROMPT" | node "$RUNNER" ... -- NEVER echo. JSON via heredoc.
• Validate: init output must start with CODEX_SESSION:. start/resume must return valid JSON. CODEX_NOT_FOUND->tell user install codex.
• status === "completed" means Codex's turn is done -- NOT that the debate is over. MUST check Loop Decision table.
• Loop: Do NOT exit unless consensus or stalemate. No round cap.
• Errors: failed->retry once (re-poll 15s). timeout->report partial, suggest lower effort. stalled+recoverable->stop->recovery resume->poll; not recoverable->report partial. Cleanup sequencing: finalize+stop ONLY after recovery resolves.
• Cleanup: ALWAYS run finalize + stop, even on failure/timeout.
• Runner manages all session state -- NEVER read/write session files manually.
• Information barrier: Claude MUST complete independent analysis BEFORE reading Codex output.
• NEVER modify code -- report + suggest only.
• For detailed error flows -> Read references/protocol.md

Workflow

1. Collect Inputs

Mode: git diff --cached --quiet exit 1=staged, exit 0=last. Effort: <=200 lines=low, 201-1000=medium, >1000=high. Staged: git diff --cached, files changed. Last: git log -n N, clamp N to history, diff. Context discovery: language/framework, linters, test frameworks, CI config.

2. Init + Render + Start (Do NOT poll yet)

Init: node "$RUNNER" init --skill-name codex-commit-review --working-dir "$PWD" Render: template=staged-round1 or last-round1. Placeholders: FILES_CHANGED, DIFF_CONTEXT, USER_REQUEST, SESSION_CONTEXT, PROJECT_CONTEXT, COMMIT_LIST (last only). Start: printf '%s' "$PROMPT" | node "$RUNNER" start "$SESSION_DIR" --effort "$EFFORT"

3. Claude Independent Analysis (BEFORE polling)

INFORMATION BARRIER: MUST NOT read Codex output. Render: template=claude-staged or claude-last. Read diff/code, write FINDING-{N} per references/claude-analysis-template.md. Last mode: Evidence MUST reference SHA+subject. Overall Assessment + Strongest Positions. COMPLETE before Step 4.

4. Poll -> Cross-Analysis -> Resume Loop

Poll + report activities. Parse review.blocks[] (id, title, severity, category, location, problem, evidence) + review.overall_assessment. Fallback: review.raw_markdown. Compare Claude FINDING-{N} vs Codex ISSUE-{N}: Agreement, Disagreement, Claude-only, Codex-only, Same Direction Different Severity. Build response: Agreements, Disagreements, New findings. Claude orchestration is authoritative -- Codex VERDICT is advisory. Render: template=staged-round2+ or last-round2+. Placeholders: SESSION_CONTEXT, PROJECT_CONTEXT, AGREED_POINTS, DISAGREED_POINTS, NEW_FINDINGS, CONTINUE_OR_CONSENSUS_OR_STALEMATE, DIFF_CONTEXT, COMMIT_LIST. Resume + back to Poll.

| # | Condition | Action | |---|-----------|--------| | 1 | Full/Partial Consensus (no severity >= medium disagreements) | EXIT -> step 5 | | 2 | convergence.stalemate === true | EXIT -> step 5 (stalemate) | | 3 | Disagreements severity >= medium remain | CONTINUE -> Cross-Analysis |

5. Completion + Output

Report: Rounds, Verdict, Claude/Codex Findings, Agreed/Disagreed counts, Files Reviewed, FINDING<->ISSUE Mapping, Overall Assessment (Code quality/Security/Test coverage/Maintainability).

6. Finalize + Cleanup

finalize + stop. Always run. (-> references/protocol.md for error handling)

Flavor Text Triggers

SKILL_START, POLL_WAITING, CODEX_RETURNED, THINK_PEER, THINK_AGREE, THINK_DISAGREE, SEND_REBUTTAL, LATE_ROUND, APPROVE_VICTORY, STALEMATE_DRAW, FINAL_SUMMARY

Rules

• Safety: NEVER git commit --amend, git rebase, or modify commit history.
• Both Claude and Codex are equal peers. For last mode N>1: reference specific commit SHA in Evidence.