codex-compat
Skill4 starsChanged 10 days agoMITView on GitHub
Generate Codex hooks.json and AGENTS.md parity surfaces from the Claude hook manifest.
In this repository117 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- accessibilitySkill
- agent-watchdogSkill
- aipSkill
- animateSkill
- animation-performanceSkill
- animation-vocabularySkill
- ask-benSkill
- ask-sonnerSkill
- bazelSkill
- bend-lawsSkill
- blast-radiusSkill
- brain-dumpSkill
- break-uiSkill
- ci-fixesSkill
- codebase-designSkill
- codexSkill
- commit-push-prSkill
- connect-querySkill
- create-verification-skillSkill
- demoSkill
- development-lifecycleSkill
- diagnosing-bugsSkill
- docsSkill
- dogfoodSkill
- domain-modelingSkill
- e2e-testingSkill
- effect-tsSkill
- effect-v3-to-v4Skill
- efficient-frontierSkill
- eli5Skill
- excalidraw-diagramSkill
- extend-harnessSkill
- frontend-invariantsSkill
- frontend-starter-kitSkill
- full-sendSkill
- gesturesSkill
- golang-reviewSkill
- golangSkill
- goSkill
- grillingSkill
- handoffSkill
- hook-auditSkill
- implement-specSkill
- improve-animationsSkill
- improve-codebase-architectureSkill
- improveSkill
- launchdarklySkill
- lie-detectorSkill
- maintain-verification-skillSkill
- make-pr-easy-to-reviewSkill
- mobile-nativeSkill
- plan-arbiterSkill
- plow-aheadSkill
- postgresqlSkill
- primeSkill
- productSkill
- protobufSkill
- prototypeSkill
- The first 60 shown.
The file
Files
Codex supports Claude-style SessionStart, SubagentStart, PreToolUse, PermissionRequest, PostToolUse, PreCompact, PostCompact, UserPromptSubmit, SubagentStop, and Stop hooks. Claude-only events (FileChanged, WorktreeCreate, SessionEnd, PostToolUseFailure) use a safe Stop-batch fallback or are dropped. Tool matchers cover Bash, MCP, apply_patch, and Edit|Write; map edit hooks direct when possible. Use /read-the-damn-docs for current behavior and /plan-arbiter for ambiguous mappings.
Creates
.codex/hooks.json: supported direct hooks..codex/hooks/codex-batch-check.sh: fallback-only checks..codex/rules/frontend-skills.rules: execpolicy floor.- Root
AGENTS.md+CLAUDE.md: shared rules. - compatibility matrix:
direct|direct with shim|fallback only|unsupported.
Steps
- Classify every
.claude/settings.jsonhook with REFERENCE.md. - Generate
.codex/hooks.json: lifecycle events direct; Bash/Edit|Write/apply_patch/mcp__.*pre/post direct; PermissionRequest direct only when scripts accept Codex payloads; omit/document unsafe unsupported cases. - Copy
scripts/codex-batch-check.shonly when fallback is required; make executable. - Copy
hooks/frontend-skills.rulesto.codex/rules/. - Generate AGENTS/CLAUDE from the reference template.
- Set
[features] hooks = trueinconfig.toml; run/hooksand trust definitions again after changes. CI uses managed hooks or--dangerously-bypass-hook-trust. Optionalnotify = ["bash", "<repo>/.claude/hooks/codex-notify.sh"].
Verify
- Direct source
Edit|WritePostToolUse hooks exist; batch checker exists only for fallback. - Rules copy is byte-identical; hooks active/trusted, not pending.
- Root AGENTS/CLAUDE exist;
.claude/settings.jsonunchanged.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.