Get Codex's review of Claude's implementation plans. Trigger when user wants a second opinion on a plan ("have Codex review this plan", "get second opinion from Codex", "critique this plan with Codex"), or after Claude creates a plan file that needs validation before implementation.
Inherits all available tools
Additional assets for this skill
This skill inherits all available tools. When active, it can use any tool Claude has access to.
Have Codex critique Claude's implementation plans for a second perspective.
cat ~/.claude/plans/example-plan.md | codex exec "Review this implementation plan:
\$(cat)
Do not make any changes. Provide critique and feedback only." --sandbox read-only --ask-for-approval never 2>&1
Claude writes plans to ~/.claude/plans/. Pipe plan content via stdin since Codex may not have access to that directory:
cat ~/.claude/plans/example-plan.md | codex exec "Review this implementation plan:
\$(cat)
Consider:
1. Are there gaps or missing steps?
2. Are there risks not addressed?
3. Is the approach optimal?
4. What alternatives should be considered?
Do not make any changes. Respond with feedback only." --sandbox read-only --ask-for-approval never 2>&1
Pipe the plan via stdin and let Codex read source files from the project:
cat ~/.claude/plans/auth-refactor.md | codex exec "Review this implementation plan:
\$(cat)
Also read these source files for context:
- src/auth/login.ts
- src/middleware/session.ts
Evaluate if the plan addresses the actual codebase structure. Do not make any changes. Respond with feedback only." --sandbox read-only --ask-for-approval never 2>&1
Risk assessment:
cat ~/.claude/plans/migration.md | codex exec "Review this plan for risks:
\$(cat)
Evaluate:
- Breaking changes
- Data loss potential
- Rollback complexity
- Dependencies that could fail
Do not make any changes. Respond with feedback only." --sandbox read-only --ask-for-approval never 2>&1
Completeness check:
cat ~/.claude/plans/feature.md | codex exec "Review this plan for completeness:
\$(cat)
Evaluate:
- Are all edge cases covered?
- Is testing addressed?
- Are there missing steps?
Do not make any changes. Respond with feedback only." --sandbox read-only --ask-for-approval never 2>&1
--dangerously-bypass-approvals-and-sandbox or --sandbox danger-full-access - these disable safety features and are forbidden$(cat) - Codex may not read ~/.claude/plans/ directly--sandbox read-only to prevent file modifications--ask-for-approval never for non-interactive executioncodex --help to verify correct flag usagedangerouslyDisableSandbox: true for Bash callsreferences/setup.md for troubleshooting