Codex
Wraps Claude around the OpenAI Codex CLI for exec/resume coding runs — needs codex installed.
Test report
- Verdict
- Works with setup
- Score
- Tested
- Jul 20, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 82f09ba
Fetched the repo tree (monorepo of plugins) and read skills/codex/SKILL.md + README raw. SKILL.md is a clean command-builder for `codex exec`/`resume --last` with a sandbox/model/reasoning-effort table; no scripts or relative links to break, no exfiltration (only benign `2>/dev/null`). Two real doc flaws: the model specs read as fabricated/unverifiable (GPT-5.2, gpt-5.2-max, "76.3% SWE-bench", pricing, Sept-2024 cutoff) and the README contradicts itself (its example asks for "gpt-5 or gpt-5-codex"); and the README's install command clones a DIFFERENT repo, git@github.com:skills-directory/skill-codex.git via SSH, not softaworks/agent-toolkit. OUTPUT could not be measured: the skill's value is the actual Codex run, which needs the external OpenAI `codex` CLI installed + authenticated — NOT measured (not a measured loss). Trigger phrasings (3 activate / 2 not): ACTIVATE — "Use codex to analyze this repo and suggest improvements", "Run codex exec to refactor the auth module", "Resume the last codex session and add error handling"; NOT — "Refactor this module for me" (generic, no codex), "Explain what a codex is in medieval manuscripts". All 5 correct.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 4/5
- Triggers reliably 5/5
- Output vs. baseline 3/10
- Docs & honesty 2/5
What Codex does
A pure-prompt skill that teaches Claude how to build and run OpenAI Codex CLI commands (codex exec, resume --last) with the right model, reasoning effort, and sandbox mode, then suppress thinking tokens and summarize output. Triggers when the user says "use codex", "codex exec", or "resume the codex session". Does nothing on its own — it requires the external codex CLI installed and authenticated.
How to install Codex
git clone --depth 1 https://github.com/softaworks/agent-toolkit /tmp/at && mkdir -p ~/.claude/skills && cp -r /tmp/at/skills/codex ~/.claude/skills/codex && rm -rf /tmp/at # then: install & authenticate the OpenAI `codex` CLI (codex --version must succeed)
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Codex
-
/codexWraps Claude around the OpenAI Codex CLI for exec/resume coding runs — needs codex installed.
It also activates on plain-language prompts like these:
-
Use Codex to refactor this module -
Run codex exec to analyze this codebase -
Resume my last Codex session on this repo
Frequently asked questions
- Is the Codex skill free?
- Yes. The skill itself is free from softaworks/agent-toolkit. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Codex work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 20, 2026. Verdict: Works with setup. Fetched the repo tree (monorepo of plugins) and read skills/codex/SKILL.md + README raw. SKILL.md is a clean command-builder for `codex exec`/`resume --last` with a sandbox/model/reasoning-effort table; no scripts or relative links to break, no exfiltration (only benign `2>/dev/null`). Two real doc flaws: the model specs read as fabricated/unverifiable (GPT-5.2, gpt-5.2-max, "76.3% SWE-bench", pricing, Sept-2024 cutoff) and the README contradicts itself (its example asks for "gpt-5 or gpt-5-codex"); and the README's install command clones a DIFFERENT repo, git@github.com:skills-directory/skill-codex.git via SSH, not softaworks/agent-toolkit. OUTPUT could not be measured: the skill's value is the actual Codex run, which needs the external OpenAI `codex` CLI installed + authenticated — NOT measured (not a measured loss). Trigger phrasings (3 activate / 2 not): ACTIVATE — "Use codex to analyze this repo and suggest improvements", "Run codex exec to refactor the auth module", "Resume the last codex session and add error handling"; NOT — "Refactor this module for me" (generic, no codex), "Explain what a codex is in medieval manuscripts". All 5 correct.
- What is the Codex SkillProof Score?
- 5.6/10 — installs cleanly 4/5, triggers reliably 5/5, output vs. baseline 3/10, docs & honesty 2/5.
- How do I install Codex?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Codex with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.