CodexDelegate
Delegate coding, repository analysis, file edits, test runs, or code review to the local Codex CLI without embedding an OpenAI API key. This skill was created to work with ChatGPT/…
jlacroix82
@jlacroix82
What This Skill Does
Delegates coding, repository analysis, file edits, test runs, or code review to the local Codex CLI without embedding an OpenAI API key. It invokes Codex exec using existing ChatGPT/Codex CLI authentication and returns Codex's final output.
Replaces embedding an OpenAI API key or writing separate integrations for each coding task by leveraging the local Codex CLI with existing authentication.
When to Use It
- Ask Codex to refactor a specific file in the repository
- Delegate a code review of a pull request to Codex
- Have Codex summarize the repository and list riskiest files
- Let Codex create or edit documentation files in the repo
- Pipe test output to Codex and ask it to fix failing tests
- Request Codex to analyze code for security or performance issues
Install
$ openclaw skills install @jlacroix82/agent-codex-delegateCodex Delegate
Use this skill when the user wants OpenClaw to hand a coding or repository task to Codex and receive a result back. It is meant for local, trusted operator setups where codex is already installed and signed in.
Preconditions
codexis installed onPATH.- Codex CLI has already been authenticated by the operator, preferably with ChatGPT sign-in. Do not ask the user for an API key.
- Run inside the target repository or pass
--cwd. - OpenClaw must allow this agent to use
execfor the wrapper script.
If Codex is not authenticated, tell the operator to run codex or codex login once and complete the ChatGPT browser login. Do not request, print, copy, or inspect ~/.codex/auth.json.
Wrapper
Prefer the bundled wrapper:
{baseDir}/scripts/codex-delegate.sh --cwd /path/to/repo --prompt "summarize the repository and list the riskiest files"
For file creation or edits:
{baseDir}/scripts/codex-delegate.sh \
--cwd /path/to/repo \
--sandbox workspace-write \
--prompt "Create docs/architecture.md explaining the current service layout."
For machine-readable event logs while still returning the final Codex answer:
{baseDir}/scripts/codex-delegate.sh \
--cwd /path/to/repo \
--sandbox workspace-write \
--json-log .openclaw/codex-runs/last.jsonl \
--output .openclaw/codex-runs/last.md \
--prompt "Review the auth changes and fix any failing tests."
To pass CLI-style context:
npm test 2>&1 | {baseDir}/scripts/codex-delegate.sh \
--cwd /path/to/repo \
--sandbox workspace-write \
--prompt "Summarize the failing tests and make the smallest safe fix."
Use --prompt-file path when a generated prompt is already on disk. Use --stdin-file path when another tool wrote context to a file.
Permissions
This skill requires:
exec: Run thecodexCLI and the bundled wrapper scriptfs.read: Read repository files for analysisfs.write: Write files when--sandbox workspace-writeis used
Scope
This skill is limited to delegating coding and repository tasks to the local Codex CLI. It does not manage Codex configuration, handle authentication, or perform actions outside the target repository. All write operations require explicit user confirmation before proceeding.
Delegation Workflow
- Restate the exact Codex task in one compact prompt.
- Choose the narrowest sandbox:
read-onlyfor analysis, review, summaries, planning.workspace-writefor creating or editing files in the repo.danger-full-accessonly on an isolated machine/container with explicit operator approval.
- Run the wrapper from the repository root or with
--cwd. - Return Codex's final answer to the user.
- If edits were allowed, inspect
git diff --statand relevant diffs before claiming files were changed.
Safety Rules
- Never pass
OPENAI_API_KEY,CODEX_API_KEY, or raw tokens to Codex. The wrapper unsets API-key env vars before invokingcodex exec. - Never read, reveal, summarize, or copy
~/.codex/auth.json. - Do not run this skill from public or untrusted chat channels.
- Do not give Codex secrets, credentials, private keys, or production data unless the operator explicitly approved that exact data flow.
- Prefer an isolated checkout or git worktree for write tasks.
- Treat Codex output as another agent's report; verify important claims locally.
Failure Handling
codex: command not found: ask the operator to install Codex CLI.- Authentication failure: ask the operator to run
codex loginwith ChatGPT sign-in, then retry. - Git repo error: run in a git repository, or use
--skip-git-repo-checkonly for a deliberately isolated scratch workspace. - Permission error on edits: retry with
--sandbox workspace-writeonly if file changes are intended.
Top skills in this category
Skill Vetter
@spclaudehomeSecurity-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
Github
@steipeteInteract with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Humanizer
@biostartechnologyRemove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.
Free Ride - Unlimited free AI
@shaivpidadiManages free AI models from OpenRouter for OpenClaw. Automatically ranks models by quality, configures fallbacks for rate-limit handling, and updates opencla...
Elite Longterm Memory
@nextfrontierbuildsUltimate AI agent memory system for Cursor, Claude, ChatGPT & Copilot. WAL protocol + vector search + git-notes + cloud backup. Never lose context again. Vibe-coding ready.