Where this fits

You're on a feature branch with a dozen commits ready for review. Before opening the PR, you want a thorough, structured critique of the changes—without spending an hour manually scanning diffs. This skill runs a review against git diff main...HEAD, returns a verdict and detailed follow-ups, and optionally posts the result to your PR or applies non-blocking suggestions directly. It turns the review loop into a faster, more consistent step in your daily workflow.

Why agents benefit

• Deterministic branching context: The skill enforces a clear precondition—must be on a non-main branch with commits ahead of main—preventing agents from reviewing stale or empty diffs.
• Actionable structured output: It emits a consistent review format (verdict, blocking issues, follow-ups) that agents can parse and act on without extra prompting.
• Optional PR integration: With --pr, the agent posts the review as a PR comment using gh, making the review a permanent part of the collaboration thread.
• Inline follow-up application: --apply-followups lets the agent edit files directly for non-blocking suggestions—no back-and-forth needed.
• Safe prompting by default: Without flags, the agent prompts before making changes, giving you control while still moving fast.

Practical scenarios

After finishing a feature branch, you ask the agent to review without any flags. It outputs a structured review locally, highlighting potential edge cases and style issues. You decide which suggestions to tackle manually.

You run with --pr --apply-followups on a branch with an open PR. The agent applies all non-blocking suggestions directly, then posts the review to the PR—leaving only blockers for you to address.

You suspect a critical flaw but aren't sure. The agent runs a review and returns a CHANGES_REQUESTED verdict. It stops, tells you the blockers, and avoids applying any follow-ups until you resolve them—keeping the branch safe.

Add it to your agent workflow

Install the skill with:

npx skills add mhawthorne/gza --skill gza-code-review-interactive

Then invoke it from your agent prompt, e.g., "Run interactive code review and post to PR with follow-ups being applied." The skill will handle the rest.

For a deeper look at the skill's capabilities and how it fits the broader agent skill ecosystem, see its listing on OpenAgentSkill.

Compare before adopting

Before trusting this skill, measure three things: quality of the review output (does it catch real bugs or just style nits?), maintenance freshness (check commit history and issue responses), and alternatives (other review skills like auto-review or manual agent prompts). Also confirm the workflow fits—it requires a Git-based setup and a gh CLI for PR posting.

Why it is worth tracking

This skill sits at a sweet spot: small, focused, and actively maintained (v2.4.0). With 11 stars and a clear scope, it's early but signals utility. If you're building or refining code review automation for agents, track this skill to see how its structured output evolves and whether its --apply-followups feature matures into a safe default. Evaluate it when your team standardizes review feedback loops or when you need a default review step in CI-integrated agent workflows.