Agentic Code Review
SkillSecurityWith this skill, your AI can review code changes for problems before they merge. It looks at a diff, pull request, branch, or AI-generated code and flags correctness, security, regression, test, performance, and maintainability risks. It works as a final check before changes are merged.
Available today. Use it from your connected AI after setup.
No other account needed.
After adding the skill, ask your AI to review a diff, pull request, or branch before you merge it. It can also review code that your AI has written.
Then ask your AI: use the Agentic Code Review skill
What your AI can do with it
- Review diffs, pull requests, and branches before merge
- Check AI-generated code for correctness and security risks
- Flag changes that could break what already worked
- Point out test risks in proposed changes
- Spot performance problems in new code
- Highlight changes that make code harder to maintain
What this skill tells your AI
The instructions your AI receives, as published by dominiktobureto/awesome-grok-build in .grok/skills/agentic-code-review/SKILL.md and read by ahel’s review.
Review like a senior engineer who wants the change to ship safely.
Grok Build Mode
- Use Plan Mode only if the review requires a fix plan. Pure review is read-only.
- Use subagents for independent passes on risky diffs:
correctness: behavior and edge cases.security: auth, permissions, secrets, injection, dependency risk.tests: coverage and missing regression tests.performance: complexity, queries, caching, bundle size.
- Arena-style review: if Arena Mode is available, compare independent review outputs and keep only findings with concrete failure paths. Otherwise synthesize subagent findings and dedupe.
- Human-in-the-loop: never auto-apply fixes unless the user explicitly asks. Findings first, fixes after approval.
Review Priority
- Correctness bugs.
- Security and privacy risks.
- Data loss, migrations, auth, payments, permissions, and concurrency.
- Missing or weak tests.
- Performance regressions.
- Developer experience and maintainability.
- Style only when it blocks comprehension or violates repo conventions.
Process
- Read the diff and the surrounding code.
- Identify the intended behavior from the issue, README, tests, or prompt.
- Check whether the implementation actually satisfies that behavior.
- Look for edge cases and failure modes.
- Verify test coverage maps to the risk.
- Run narrow verification only if the user asked for fix validation or the repo has cheap checks.
- Produce findings first. Keep summary short.
Finding Format
Use this format:
[P1] Title
File: path/to/file.ext:line
Problem: ...
Why it matters: ...
Suggested fix: ...
Severity:
P0: must fix before merge, production-breaking or data/security critical.P1: should fix before merge.P2: useful fix, not necessarily blocking.P3: polish or follow-up.
Example Prompts
Use agentic-code-review on the current diff. Findings first. Ignore style unless it creates real risk.
Review this PR with subagents for correctness, security, tests, and performance. Dedupe findings and keep only actionable issues.
Git Best Practices
- Review staged and unstaged changes separately when possible.
- Mention untracked files that affect the change.
- Do not recommend committing until verification is complete.
- If proposing fixes, keep each fix scoped to one finding.
Guardrails
- Do not list style nits as findings unless they create real risk.
- Do not claim a bug without a concrete failure path.
- If the diff is safe, say so and name residual risks or missing verification.
- Prefer references to exact files and lines.
Signals
- GitHub stars
- 62
- Forks
- 14
- Last commit
- Aug 2026
Advanced
- Catalog kind
- skill
- Gateway key
agentic-code-review- Source
- github.com/dominiktobureto/awesome-grok-build