traiage-review

Prioritizes changed repository files that require human code review attention.

1.5k|154|Updated Nov 13, 2025
One-click install
npx skills add https://github.com/NeoLabHQ/context-engineering-kit --skill traiage-review
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: traiage-review
Source: https://github.com/NeoLabHQ/context-engineering-kit/tree/main/plugins/review/skills/traiage-review
Command: npx skills add https://github.com/NeoLabHQ/context-engineering-kit --skill traiage-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Human reviewers have limited time and cannot inspect every changed file in a large pull request or branch. This Skill builds a focused list of the files most likely to cause a review to fail, so reviewers spend attention where it matters.

Core Features & Use Cases

  • Multi-Agent Triage: Launches four parallel agents (change-story, change-impact, change-failure, change-expectation) that each rank files by importance, severity, detectability, and confidence.
  • Three Review Modes: Supports local-changes (staged/unstaged/untracked), branch-diff against the detected default branch, and single-commit review workflows.
  • Random Sampling: Adds a random sample of changed files via a Python script to catch issues the agents may have missed, plus a separate list of declarative files.
  • Use Case: Before reviewing a 200-file pull request, run this Skill to get a ranked table of roughly 25 key files with key facts, risks, and design decisions summarized.

Quick Start

Ask the agent to triage the current changes for review, for example: prioritize which changed files in this branch I should review first.

Frequently Asked Questions about traiage-review

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

FAQPage Schema
How do I prioritize files for code review in a large pull request?

Run this Skill on the branch or changeset; it launches four parallel agents that each rank changed files by importance, severity, detectability, and confidence, then merges the top candidates into a final list of about 20 key files plus a random sample.

How to triage code changes before human review with git?

The Skill detects the review mode automatically: local staged/unstaged/untracked changes, a branch diff against the detected default branch, or a specific commit. It uses read-only git commands like git status, git diff, and git show to gather the changed file set.

Can I review a specific commit instead of local changes?

Yes. Pass a commit parameter with a concrete hash, or pass 'commit' with no hash to resolve the latest commit via git rev-parse HEAD. The Skill then reviews the diff introduced by that commit using git show or git diff.

Does the triage skill modify my repository or source files?

No. The orchestrator is restricted to read-only git commands and the random sample script. It explicitly forbids reading source files or running any mutating git command such as commit, checkout, reset, or rebase.

What are the limitations of automated review triage?

The output is a prioritization heuristic, not a full review; it selects roughly 20 agent-ranked files plus 5 random samples, so low-ranked files may still contain issues. It also depends on the quality of the four sub-agents' scoring and confidence ratings.