| name | codex-cli |
| description | Use Codex CLI for web search, code review, and writing review. Triggers on (1) web search requests, (2) code review of git changes, (3) writing/file review. Use for "search the web for...", "review this code", "review my changes", "review this file". |
Codex CLI
OpenAI's Codex CLI for web search and reviews. All commands are non-interactive.
Web Search
ChatGPT-tuned agentic search. Default: gpt-5.2 with medium reasoning.
codex exec "your question" --enable web_search_request -m gpt-5.2 -c model_reasoning_effort="medium" 2>/dev/null
Note: 2>/dev/null suppresses metadata and reasoning traces, returning only the answer.
Code Review (Git Changes)
Default: gpt-5.2-codex with high reasoning.
# Review uncommitted changes (staged + unstaged + untracked)
codex review --uncommitted -c model="gpt-5.2-codex" -c model_reasoning_effort="high"
# Review against a base branch
codex review --base main -c model="gpt-5.2-codex" -c model_reasoning_effort="high"
# Review a specific commit
codex review --commit <sha> -c model="gpt-5.2-codex" -c model_reasoning_effort="high"
# Custom review instructions
codex review --uncommitted "Focus on security issues" -c model="gpt-5.2-codex" -c model_reasoning_effort="high"
Writing/File Review
Review a file for clarity, density, and fluff removal.
codex exec "Review the writing at <filepath>. Provide an improved version in full, then explain each change. Prioritize: removing fluff, increasing information density, improving clarity. No stylistic preferences." --skip-git-repo-check 2>/dev/null
Use your judgment to incorporate high-value changes directly. Briefly summarize what you changed and why.
Reasoning Effort
Override with -c model_reasoning_effort="low|medium|high". Higher = deeper analysis, slower.