kreview

Triages PR review comments and implements, pushes back, or escalates each finding.

2|1|Updated Feb 5, 2026
One-click install
npx skills add https://github.com/kpiteira/devops-ai --skill kreview-kpiteira
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: kreview
Source: https://github.com/kpiteira/devops-ai/tree/main/skills/kreview
Command: npx skills add https://github.com/kpiteira/devops-ai --skill kreview-kpiteira

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Pull request review rounds generate noisy feedback from bots and humans, and blindly implementing every comment wastes effort or degrades the code. This Skill critically assesses each review comment against the PR's declared review scope, decides whether to implement, push back, discuss, or file it as out of scope, and executes the decision with a full audit trail. ## Core Features & Use Cases - Complete feedback fetch: Pulls reviews, review threads, issue comments, CI state, and Copilot's suppressed comments via the GitHub CLI so no finding is missed. - Critical triage: Classifies each finding as IMPLEMENT, PUSH BACK, DISCUSS, or OUT OF SCOPE, with an isolated-vs-systemic root-cause analysis and commit-level provenance (original diff vs review-fix commit). - Attended and autonomous modes: Runs interactively with a confirmation table, or autonomously when driven by the kbabysit multi-round loop. - Round report: Produces a structured per-round report with verdicts, provenance totals, gate status, and items escalated for the human. - Use Case: After pushing code, Copilot leaves 15 findings across a PR. The Skill fetches all of them, implements the two real bugs, pushes back on style nitpicks with reasoning, files one out-of-scope defect as an issue, and posts a round report. ## Quick Start Ask the assistant to run kreview on the current pull request to triage and address all outstanding review comments.

Frequently Asked Questions about kreview

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

FAQPage Schema
How do I address GitHub PR review comments automatically?

Run the kreview skill on the pull request. It fetches all reviews, threads, issue comments, and CI state via the GitHub CLI, triages each finding as implement, push back, discuss, or out of scope, then applies fixes, replies to every thread, and posts a round report.

How to triage Copilot review comments on a pull request?

The skill parses Copilot's review bodies including the suppressed comments section, which often contains findings that never open threads. Each finding is assessed against the PR's review scope, and low-value suggestions are pushed back with reasoning rather than blindly implemented.

Does this work with reviewers other than Copilot?

Yes, it works with any reviewer including humans and other bots. When multiple reviewers flag the same issue it is treated as high-confidence signal, and contradictory comments are escalated for human judgment.

What happens to real bugs that are outside the PR's scope?

Out-of-scope findings are never committed to the branch. The skill files a GitHub issue containing the finding verbatim with a link back to the thread, replies to the reviewer naming the scope outcome it does not serve, and resolves the thread.

When should I use kreview versus kbabysit?

Use kreview to process a single round of review feedback on a PR. Use kbabysit when you want a multi-round loop that repeatedly invokes kreview, requests re-reviews, and drives the PR to merge-ready state.