codex
Use Cautionnpx machina-cli add skill softaworks/agent-toolkit/codex --openclawFiles (1)
SKILL.md
5.1 KB
Codex Skill Guide
Running a Task
- Default to
gpt-5.2model. Ask the user (viaAskUserQuestion) which reasoning effort to use (xhigh,high,medium, orlow). User can override model if needed (see Model Options below). - Select the sandbox mode required for the task; default to
--sandbox read-onlyunless edits or network access are necessary. - Assemble the command with the appropriate options:
-m, --model <MODEL>--config model_reasoning_effort="<high|medium|low>"--sandbox <read-only|workspace-write|danger-full-access>--full-auto-C, --cd <DIR>--skip-git-repo-check
- Always use --skip-git-repo-check.
- When continuing a previous session, use
codex exec --skip-git-repo-check resume --lastvia 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/nullto allcodex execcommands to suppress thinking tokens (stderr). Only show stderr if the user explicitly requests to see thinking tokens or if debugging is needed. - 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 |
Model Options
| Model | Best for | Context window | Key features |
|---|---|---|---|
gpt-5.2-max | Max model: Ultra-complex reasoning, deep problem analysis | 400K input / 128K output | 76.3% SWE-bench, adaptive reasoning, $1.25/$10.00 |
gpt-5.2 ⭐ | Flagship model: Software engineering, agentic coding workflows | 400K input / 128K output | 76.3% SWE-bench, adaptive reasoning, $1.25/$10.00 |
gpt-5.2-mini | Cost-efficient coding (4x more usage allowance) | 400K input / 128K output | Near SOTA performance, $0.25/$2.00 |
gpt-5.1-thinking | Ultra-complex reasoning, deep problem analysis | 400K input / 128K output | Adaptive thinking depth, runs 2x slower on hardest tasks |
GPT-5.2 Advantages: 76.3% SWE-bench (vs 72.8% GPT-5), 30% faster on average tasks, better tool handling, reduced hallucinations, improved code quality. Knowledge cutoff: September 30, 2024.
Reasoning Effort Levels:
xhigh- Ultra-complex tasks (deep problem analysis, complex reasoning, deep understanding of the problem)high- Complex tasks (refactoring, architecture, security analysis, performance optimization)medium- Standard tasks (refactoring, code organization, feature additions, bug fixes)low- Simple tasks (quick fixes, simple changes, code formatting, documentation)
Cached Input Discount: 90% off ($0.125/M tokens) for repeated context, cache lasts up to 24 hours.
Following Up
- After every
codexcommand, immediately useAskUserQuestionto confirm next steps, collect clarifications, or decide whether to resume withcodex 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.
Error Handling
- Stop and report failures whenever
codex --versionor acodex execcommand 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.
CLI Version
Requires Codex CLI v0.57.0 or later for GPT-5.2 model support. The CLI defaults to gpt-5.2 on macOS/Linux and gpt-5.2 on Windows. Check version: codex --version
Use /model slash command within a Codex session to switch models, or configure default in ~/.codex/config.toml.
Source
git clone https://github.com/softaworks/agent-toolkit/blob/main/skills/codex/SKILL.mdView on GitHub Overview
Codex is a CLI tool to run OpenAI Codex for code analysis, refactoring, and automated edits via codex exec and codex resume. It defaults to the GPT-5.2 model for state-of-the-art software engineering and lets you tune reasoning effort and sandbox mode. It also supports resuming ongoing Codex sessions and precise output capture.
How This Skill Works
You start with a default GPT-5.2 model and optionally pick a reasoning effort level (xhigh, high, medium, or low). Then choose a sandbox mode (read-only by default; workspace-write for edits; danger-full-access for network access). The command is assembled with -m/--model, --config model_reasoning_effort, --sandbox, --full-auto, -C/--cd, and --skip-git-repo-check. Always include --skip-git-repo-check and append 2>/dev/null to suppress thinking tokens, unless you request to see them. After execution, Codex summarizes the outcome; you can resume later with codex resume using the provided syntax.
When to Use It
- Read-only review or analysis of a codebase to identify issues or opportunities for refactoring.
- Apply local edits or automated changes across files using workspace-write sandbox and --full-auto.
- Permit network access or broad access when dependencies or external data are needed during analysis.
- Resume a recent Codex session to continue analysis or edits from where you left off.
- Run Codex from another directory using -C <DIR> to target a specific project path.
Quick Start
- Step 1: Default to gpt-5.2 and ask the user which reasoning effort to use (xhigh, high, medium, low).
- Step 2: Choose the sandbox mode (read-only by default; switch to workspace-write or danger-full-access if edits or network are needed).
- Step 3: Run codex exec with -m/--model, --config, --sandbox, --full-auto, -C/--cd, and --skip-git-repo-check; append 2>/dev/null; summarize output and offer to resume with codex resume.
Best Practices
- Default to the gpt-5.2 model and explicitly ask for the reasoning effort level (xhigh, high, medium, or low) when starting a task.
- Select the sandbox mode that matches the task: read-only for analysis, workspace-write for edits, or danger-full-access for network-enabled work.
- Always include --skip-git-repo-check in codex exec commands to satisfy safety checks and consistency.
- When resuming a session, use the dedicated resume syntax and avoid adding unnecessary flags unless requested.
- Append 2>/dev/null by default to suppress thinking tokens; reveal them only if the user explicitly asks to debug.
Example Use Cases
- A developer asks Codex to perform a static analysis on a Python module to identify refactoring opportunities.
- You need to apply naming convention fixes across multiple JavaScript files using automated edits.
- The task requires fetching dependencies from the network, so you switch to danger-full-access sandbox with full-auto.
- You want to continue a previous Codex session to refine a refactor and test new changes.
- You start Codex from a different project directory to target a specific repository.
Frequently Asked Questions
Add this skill to your agents
