Get an external code review from another AI model (Codex or Gemini CLI). Use when completing a feature, bug fix, or refactor before marking work done. Also use when the user asks for a "review", "second opinion", "sanity check", or says "check with codex/gemini", or when uncertain about an approach.
How this skill is triggered — by the user, by Claude, or both
Slash command
/external-agent-review:external-agent-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Get a second opinion on code changes by calling an external reviewer CLI.
Get a second opinion on code changes by calling an external reviewer CLI.
Always run reviewer CLIs with a 10-minute timeout — detailed reviews regularly exceed the default
2-minute limit. Pass timeout: 600000 when calling the Bash tool.
codex exec "<prompt>"
gemini -p "<prompt>"
After completing a significant feature, bug fix, or refactor, ask the user if they'd like an external review before marking the work done. Don't automatically trigger a review for small changes like config tweaks, typo fixes, or single-line changes.
Ask something like: "Want me to get a Codex/Gemini review before we wrap this up?"
If the user says no, move on. If they say yes, proceed with the review.
Skip asking and just do it when:
Both reviewers have full repo access and can run git commands, read files, and explore the codebase on their own. Describe what to review in natural language — don't pre-compute diffs. Let the reviewer decide how to inspect.
npx claudepluginhub verygoodsoftwareorg/claude-plugins --plugin external-agent-reviewRuns code reviews via external LLMs (OpenAI Codex CLI, Google Gemini CLI) on uncommitted changes, branch diffs, or commits using git and bash.
Invoke external LLMs like OpenAI Codex for code reviews on git diffs with project context from CLAUDE.md and lessons. Use for second opinions with custom focus areas.
Cross-model review using OpenAI Codex to independently verify plans or code diffs, iterating up to 5 rounds. Useful for architecture decisions, non-trivial refactors, and critical config changes.