Codex Skill Guide
Running a Task
- Ask the user (via
AskUserQuestion) which model to run (gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.3-codex-spark, or gpt-5.3-codex) AND which reasoning effort to use (xhigh, high, medium, or low) in a single prompt with two questions.
- Select the sandbox mode required for the task; default to
--sandbox read-only unless edits or network access are necessary.
- Assemble the command with the appropriate options:
-m, --model <MODEL>
--config model_reasoning_effort="<xhigh|high|medium|low>"
--sandbox <read-only|workspace-write|danger-full-access>
--full-auto
-C, --cd <DIR>
--skip-git-repo-check
"your prompt here" (as final positional argument)
- Always use --skip-git-repo-check.
- When continuing a previous session, use
codex exec --skip-git-repo-check resume --last via stdin. When resuming don't use any configuration flags unless explicitly requested by the user e.g. if he species the model or the reasoning effort when requesting to resume a session. Resume syntax: echo "your prompt here" | codex exec --skip-git-repo-check resume --last 2>/dev/null. All flags have to be inserted between exec and resume.
- IMPORTANT: By default, append
2>/dev/null to all codex exec commands to suppress thinking tokens (stderr). Only show stderr if the user explicitly requests to see thinking tokens or if debugging is needed.
- IMPORTANT (stdin):
codex exec always reads stdin and concatenates it with the positional prompt -- even when the prompt is fully supplied as a positional argument. If stdin is not closed, codex blocks forever. When invoking from a harness (background tasks, hooks, scripts where stdin is not a TTY but also not closed), explicitly redirect stdin: append </dev/null to the command, e.g. codex exec ... "prompt" </dev/null 2>/dev/null. Symptom of getting this wrong: zero bytes of stdout, zero CPU accumulated, process appears hung indefinitely.
- Run the command, capture stdout/stderr (filtered as appropriate), and summarize the outcome for the user.
- After Codex completes, inform the user: "You can resume this Codex session at any time by saying 'codex resume' or asking me to continue with additional analysis or changes."
Quick Reference
| Use case |
Sandbox mode |
Key flags |
| Read-only review or analysis |
read-only |
--sandbox read-only 2>/dev/null |
| Apply local edits |
workspace-write |
--sandbox workspace-write --full-auto 2>/dev/null |
| Permit network or broad access |
danger-full-access |
--sandbox danger-full-access --full-auto 2>/dev/null |
| Resume recent session |
Inherited from original |
echo "prompt" | codex exec --skip-git-repo-check resume --last 2>/dev/null (no flags allowed) |
| Run from another directory |
Match task needs |
-C <DIR> plus other flags 2>/dev/null |
Execution timeouts
Codex produces no intermediate output — it writes the result only at completion. If the process is killed before finishing, the output file is silently empty (no error).
Preferred approach: run synchronously — eliminates timeout risk entirely and the conversation waits for the result anyway.
If running in background, set the execution timeout based on reasoning effort:
| Reasoning effort |
Timeout |
low |
150s |
medium |
300s |
high |
600s |
xhigh |
1200s |
Following Up
- After every
codex command, immediately use AskUserQuestion to confirm next steps, collect clarifications, or decide whether to resume with codex exec resume --last.
- When resuming, pipe the new prompt via stdin:
echo "new prompt" | codex exec resume --last 2>/dev/null. The resumed session automatically uses the same model, reasoning effort, and sandbox mode from the original session.
- Restate the chosen model, reasoning effort, and sandbox mode when proposing follow-up actions.
Critical Evaluation of Codex Output
Codex is powered by OpenAI models with their own knowledge cutoffs and limitations. Treat Codex as a colleague, not an authority.
Guidelines
- Trust your own knowledge when confident. If Codex claims something you know is incorrect, push back directly.
- Research disagreements using WebSearch or documentation before accepting Codex's claims. Share findings with Codex via resume if needed.
- Remember knowledge cutoffs - Codex may not know about recent releases, APIs, or changes that occurred after its training data.
- Don't defer blindly - Codex can be wrong. Evaluate its suggestions critically, especially regarding:
- Model names and capabilities
- Recent library versions or API changes
- Best practices that may have evolved
When Codex is Wrong
- State your disagreement clearly to the user
- Provide evidence (your own knowledge, web search, docs)
- Optionally resume the Codex session to discuss the disagreement. Identify yourself as Claude so Codex knows it's a peer AI discussion. Use your actual model name (e.g., the model you are currently running as) instead of a hardcoded name:
echo "This is Claude (<your current model name>) following up. I disagree with [X] because [evidence]. What's your take on this?" | codex exec --skip-git-repo-check resume --last 2>/dev/null
- Frame disagreements as discussions, not corrections - either AI could be wrong
- Let the user decide how to proceed if there's genuine ambiguity
Error Handling
- Stop and report failures whenever
codex --version or a codex exec command exits non-zero; request direction before retrying.
- Before you use high-impact flags (
--full-auto, --sandbox danger-full-access, --skip-git-repo-check) ask the user for permission using AskUserQuestion unless it was already given.
- When output includes warnings or partial results, summarize them and ask how to adjust using
AskUserQuestion.
1---2name: codex-23description: Use when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis, refactoring, or automated editing4metadata:5 source: https://skillmd.com/skills/zuoguyoupan2023/codex-26---78# Codex Skill Guide910## Running a Task111. Ask the user (via `AskUserQuestion`) which model to run (`gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.3-codex-spark`, or `gpt-5.3-codex`) AND which reasoning effort to use (`xhigh`, `high`, `medium`, or `low`) in a **single prompt with two questions**.122. Select the sandbox mode required for the task; default to `--sandbox read-only` unless edits or network access are necessary.133. Assemble the command with the appropriate options:14 - `-m, --model <MODEL>`15 - `--config model_reasoning_effort="<xhigh|high|medium|low>"`16 - `--sandbox <read-only|workspace-write|danger-full-access>`17 - `--full-auto`18 - `-C, --cd <DIR>`19 - `--skip-git-repo-check`20 - `"your prompt here"` (as final positional argument)214. Always use --skip-git-repo-check.225. When continuing a previous session, use `codex exec --skip-git-repo-check resume --last` via stdin. When resuming don't use any configuration flags unless explicitly requested by the user e.g. if he species the model or the reasoning effort when requesting to resume a session. Resume syntax: `echo "your prompt here" | codex exec --skip-git-repo-check resume --last 2>/dev/null`. All flags have to be inserted between exec and resume.236. **IMPORTANT**: By default, append `2>/dev/null` to all `codex exec` commands to suppress thinking tokens (stderr). Only show stderr if the user explicitly requests to see thinking tokens or if debugging is needed.247. **IMPORTANT (stdin)**: `codex exec` always reads stdin and concatenates it with the positional prompt -- even when the prompt is fully supplied as a positional argument. If stdin is not closed, codex blocks forever. When invoking from a harness (background tasks, hooks, scripts where stdin is not a TTY but also not closed), explicitly redirect stdin: append `</dev/null` to the command, e.g. `codex exec ... "prompt" </dev/null 2>/dev/null`. Symptom of getting this wrong: zero bytes of stdout, zero CPU accumulated, process appears hung indefinitely.258. Run the command, capture stdout/stderr (filtered as appropriate), and summarize the outcome for the user.269. **After Codex completes**, inform the user: "You can resume this Codex session at any time by saying 'codex resume' or asking me to continue with additional analysis or changes."2728### Quick Reference29| Use case | Sandbox mode | Key flags |30| --- | --- | --- |31| Read-only review or analysis | `read-only` | `--sandbox read-only 2>/dev/null` |32| Apply local edits | `workspace-write` | `--sandbox workspace-write --full-auto 2>/dev/null` |33| Permit network or broad access | `danger-full-access` | `--sandbox danger-full-access --full-auto 2>/dev/null` |34| Resume recent session | Inherited from original | `echo "prompt" \| codex exec --skip-git-repo-check resume --last 2>/dev/null` (no flags allowed) |35| Run from another directory | Match task needs | `-C <DIR>` plus other flags `2>/dev/null` |3637## Execution timeouts3839Codex produces **no intermediate output** — it writes the result only at completion. If the process is killed before finishing, the output file is silently empty (no error).4041**Preferred approach:** run synchronously — eliminates timeout risk entirely and the conversation waits for the result anyway.4243**If running in background**, set the execution timeout based on reasoning effort:4445| Reasoning effort | Timeout |46|---|---|47| `low` | 150s |48| `medium` | 300s |49| `high` | 600s |50| `xhigh` | 1200s |5152## Following Up53- After every `codex` command, immediately use `AskUserQuestion` to confirm next steps, collect clarifications, or decide whether to resume with `codex exec resume --last`.54- When resuming, pipe the new prompt via stdin: `echo "new prompt" | codex exec resume --last 2>/dev/null`. The resumed session automatically uses the same model, reasoning effort, and sandbox mode from the original session.55- Restate the chosen model, reasoning effort, and sandbox mode when proposing follow-up actions.5657## Critical Evaluation of Codex Output5859Codex is powered by OpenAI models with their own knowledge cutoffs and limitations. Treat Codex as a **colleague, not an authority**.6061### Guidelines62- **Trust your own knowledge** when confident. If Codex claims something you know is incorrect, push back directly.63- **Research disagreements** using WebSearch or documentation before accepting Codex's claims. Share findings with Codex via resume if needed.64- **Remember knowledge cutoffs** - Codex may not know about recent releases, APIs, or changes that occurred after its training data.65- **Don't defer blindly** - Codex can be wrong. Evaluate its suggestions critically, especially regarding:66 - Model names and capabilities67 - Recent library versions or API changes68 - Best practices that may have evolved6970### When Codex is Wrong711. State your disagreement clearly to the user722. Provide evidence (your own knowledge, web search, docs)733. Optionally resume the Codex session to discuss the disagreement. **Identify yourself as Claude** so Codex knows it's a peer AI discussion. Use your actual model name (e.g., the model you are currently running as) instead of a hardcoded name:74 ```bash75 echo "This is Claude (<your current model name>) following up. I disagree with [X] because [evidence]. What's your take on this?" | codex exec --skip-git-repo-check resume --last 2>/dev/null76 ```774. Frame disagreements as discussions, not corrections - either AI could be wrong785. Let the user decide how to proceed if there's genuine ambiguity7980## Error Handling81- Stop and report failures whenever `codex --version` or a `codex exec` command exits non-zero; request direction before retrying.82- Before you use high-impact flags (`--full-auto`, `--sandbox danger-full-access`, `--skip-git-repo-check`) ask the user for permission using AskUserQuestion unless it was already given.83- When output includes warnings or partial results, summarize them and ask how to adjust using `AskUserQuestion`.