From cli-agent-skills
This skill should be used when the user asks to "review staged changes with Gemini", "run a gemini-review", "have Gemini review my code", "gemini distinguished review", "gemini sre review", "gemini team review", or wants a second-opinion code review of staged git changes delegated to Gemini CLI. Requires a persona argument: distinguished, sre, or team.
How this skill is triggered — by the user, by Claude, or both
Slash command
/cli-agent-skills:gemini-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Delegate code review of staged git changes to Gemini CLI using a role-specific critique lens. Provides a second opinion from a different model compared to `codex-review`.
Delegate code review of staged git changes to Gemini CLI using a role-specific critique lens. Provides a second opinion from a different model compared to codex-review.
| Argument | Reviewer Voice |
|---|---|
distinguished | Distinguished Engineer — architectural depth, long-term maintainability, systems thinking |
sre | Mid-level SRE/DevOps Engineer — operational concerns, deployment risk, observability, toil reduction |
team | Five named reviewers in separate sections: Senior Security Engineer, Junior Security Engineer, Staff SRE, Mid-level Reliability Engineer, Principal DevOps Engineer |
When this skill triggers, first confirm staged changes exist, then run:
"${CLAUDE_PLUGIN_ROOT}/skills/gemini-review/scripts/gemini-review.sh" "<persona>"
Valid persona values: distinguished, sre, team.
If the user hasn't specified a persona, ask which one they want before running the script.
git diff --staged is empty)gemini -p~/.claude/cli-reviews/Present the markdown output the script prints to stdout. Do not add your own code review on top of it. If the user wants your opinion afterward, they will ask.
If the script exits non-zero, report the error message it printed.
Each run writes one JSON file to ~/.claude/cli-reviews/gemini-review-<persona>-<timestamp>.json. Mention the artifact path at the end of your response.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub mrjamescollins/claude-plugins --plugin cli-agent-skills