meta-codereview-current-diff

Reviews the uncommitted git diff with three parallel reviewers and arbitrates a single verdict.

6.8k|540|Updated May 6, 2026
One-click install
npx skills add https://github.com/opensquilla/opensquilla --skill meta-codereview-current-diff
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: meta-codereview-current-diff
Source: https://github.com/opensquilla/opensquilla/tree/main/src/opensquilla/skills/exp/meta-codereview-current-diff
Command: npx skills add https://github.com/opensquilla/opensquilla --skill meta-codereview-current-diff

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Single-reviewer code review loops often miss issues or drift into unfocused opinions. This Skill runs three independent reviewers over your current uncommitted diff and merges their findings into one strict verdict before you commit.

Core Features & Use Cases

  • Three Parallel Reviewers: A safety reviewer checks for SQL injection, shell injection, hardcoded credentials, and unescaped templating; a test-coverage reviewer flags new public functions or bug fixes without tests; a style reviewer flags a fixed list of anti-patterns like bare except clauses and print() calls.
  • Strict Arbitration: A final step applies priority rules to emit exactly one verdict: BLOCK, BLOCK_WITH_OVERRIDE, or PASS_WITH_NOTES.
  • Graceful Fallback: If no diff exists, reviewers return clean verdicts and the result is PASS_WITH_NOTES: clean.
  • Use Case: Before committing a bug fix, trigger the multi-reviewer diff review to confirm the change introduces no injection risks and includes a regression test.

Quick Start

Ask the agent to run a multi-reviewer diff review on your current uncommitted changes before committing.

Frequently Asked Questions about meta-codereview-current-diff

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review my uncommitted git diff before committing?

Trigger the multi-reviewer diff review, which reads the staged diff via git diff --cached HEAD and falls back to git diff HEAD if nothing is staged. Three reviewers then analyze the diff in parallel and an arbitration step returns a single verdict.

What does the multi-reviewer code review check for?

It checks three areas: safety issues like SQL injection, shell injection, and hardcoded credentials; missing tests for new public functions or bug fixes; and style anti-patterns such as bare except clauses, print() calls, and functions over 80 lines.

What do the BLOCK and PASS_WITH_NOTES verdicts mean?

BLOCK means a critical safety issue was found and the change should not be committed. BLOCK_WITH_OVERRIDE means warnings or missing tests exist but the user may proceed with explicit acknowledgement. PASS_WITH_NOTES means the diff is acceptable, with optional style notes attached.

What happens if there is no diff to review?

If the diff read step returns NO_DIFF, all three reviewers return their clean verdicts and the arbitration step emits PASS_WITH_NOTES: clean, so the workflow completes without errors.

What happens if one of the reviewer steps fails?

If a reviewer LLM step fails, the orchestrator's partial outputs remain visible in step_outputs. The operator should manually re-run the failed reviewer rather than relying on an incomplete verdict.