What happens if new comments appear after I push a fix?

The skill features 'kickback logic' which automatically moves the workflow back to the UNDERSTAND phase if new comments are detected, ensuring that iterative feedback is handled with the same rigor.

Can I use this for any git repository?

Yes, this skill is designed for the Claude Code CLI and uses standardized shell commands, making it compatible with any local repository undergoing code review.

Why are my edit tools blocked during the workflow?

Edit and Write tools are intentionally blocked during the initial phase to force a deep understanding of the reviewer's concern before any code is modified, preventing 'cargo cult' fixes.

How does the PR Comment Workflow prevent incorrect fixes?

It implements a mandatory 'UNDERSTAND' phase that blocks code editing tools until the agent articulates the problem and the reviewer's goal in its own words, preventing mindless compliance.

What are Adversary Gates in this skill?

Adversary Gates are validation steps where the system challenges the agent's logic to ensure the proposed fix aligns perfectly with the reviewer's feedback without unintended side effects.

PR Comment Workflow

Name: PR Comment Workflow
Author: c-daly

byc-daly

0•

Herramientas de Colaboración

Streamlines pull request feedback resolution by enforcing an understanding-first workflow that prevents superficial fixes.

The PR Comment Workflow skill is a specialized enforcement system for Claude Code designed to improve the quality of code review resolutions. It guides the AI agent through a rigorous multi-phase sequence, starting with a mandatory understanding phase where the agent must articulate the reviewer's concern before any code changes are permitted. By blocking editing tools during the initial phase and utilizing adversary gates to validate logic and implementation, this skill eliminates 'cargo cult' fixes and scope creep, ensuring that every PR comment is addressed with intentionality and verified against existing tests.

Características Principales

01Kickback Logic: Automatically returns the agent to the understanding phase if new review comments are detected during the process.

02Understanding-First Enforcement: Blocks code editing until the reviewer's intent is clearly articulated and recorded.

03Phase-Based Permission Control: Dynamically restricts tool access (Read, Edit, Bash) based on the current stage of the workflow.

04Automated Verification Loop: Requires successful test and lint passes before allowing changes to be pushed.

050 GitHub stars

06Adversary Gates: Validates the agent's logic to ensure the fix matches the expressed concern without unnecessary changes.

Casos de Uso

01Preventing AI agents from making 'extra' changes or refactors that weren't specifically requested by the human reviewer.

02Standardizing the PR resolution process across a team using agentic workflows to ensure consistent quality and focus.

03Resolving complex code review feedback where architectural intent is more important than simple syntax changes.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add c-daly/agent-swarm pr-comment

For use in Claude.ai and ChatGPT

Download Skill