review-loop

Automate a three-round review, fix, verify, and re-review cycle.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/hack-ink/codexlab --skill review-loop-hack-ink
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: review-loop
Source: https://github.com/hack-ink/codexlab/tree/main/.codex/skills/review-loop
Command: npx skills add https://github.com/hack-ink/codexlab --skill review-loop-hack-ink

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Automates a bounded, repeatable review cycle that validates changes against a concrete diff or repaired branch state, ensuring ownership of findings and formal escalation when needed.

Core Features & Use Cases

  • Defines a three-round escalation: review -> fix -> verify -> re-review, with binding to the head SHA.
  • Outputs a machine-readable envelope with status, head_sha, and evidence for each reviewed state.
  • Prevents blind acceptance of changes and ensures traceable, auditable review across coordinated teams.

Quick Start

Apply the bounded review loop to the current diff and iteratively fix, verify, and re-review until the head state is clean or escalated.

Frequently Asked Questions about review-loop

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

FAQPage Schema
How do I automate a code review workflow with a bounded fix and verify loop?

To automate a code review workflow, this Skill applies a bounded three-round loop of review, fix, verify, and re-review to validate diffs, escalating to a research plugin when needed.

What is the best way to ensure traceable reviews bound to a specific branch commit SHA?

Ensuring traceable reviews requires binding decisions to the reviewed head commit SHA, which this Skill does by outputting a machine-readable envelope with status, head_sha, and evidence for each reviewed state.

How does the three-round escalation process work for validating repaired branch states?

The three-round escalation process works by iterating through review, fix, verify, and re-review cycles on a repaired branch state, escalating to the research plugin only if the head state remains unresolved.

Can I prevent blind acceptance of diffs in my automated verification workflow?

Yes, you can prevent blind acceptance of diffs by enforcing a bounded, automated review loop that requires explicit verification and generates machine-readable evidence bound to the head SHA.

When should I escalate a diff review to a research plugin instead of continuing the loop?

You should escalate a diff review to a research plugin when the bounded three-round cycle of review, fix, and verify fails to produce a clean head state for the current branch.