code-review

Reviews pull request diffs against specs and repository contracts for correctness and regressions.

1|Updated Sep 15, 2026
One-click install
npx skills add https://github.com/amoai-tech/mdeai --skill code-review-amoai-tech
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/amoai-tech/mdeai/tree/main/.claude/skills/code-review
Command: npx skills add https://github.com/amoai-tech/mdeai --skill code-review-amoai-tech

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Reviewing a PR, branch, or work-in-progress diff requires checking both spec fit and code quality against the exact current head, which is easy to get wrong when reviewers rely on stale assumptions or unverified bot comments. ## Core Features & Use Cases - Two-axis review: Maps each acceptance criterion to code, tests, and evidence while checking correctness, security, data integrity, and maintainability. - Structured findings: Every actionable finding includes severity, category, file:line evidence, impact, smallest safe fix, and a verification probe. - Bot claim verification: Disproves or confirms reviewer/bot comments against exact-head code, marking disproven claims as NOISE. - Use Case: Given a branch and a Linear task like SAN-123, pin the merge-base and head SHA, then produce a review that flags missing acceptance criteria and material risks without style trivia. ## Quick Start Ask the assistant to review this branch against main and the linked task spec, citing current-head evidence for every finding.

Frequently Asked Questions about code-review

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

FAQPage Schema
How do I review a pull request against a task spec?

Pin the exact head SHA and merge-base, then map each acceptance criterion from the task to code, tests, and evidence. Flag missing behavior, invented scope, or implementations that do not produce the requested user outcome.

How to verify automated code review bot comments?

Check each bot claim against the exact current-head code before acting on it. If the claim is disproven by current-head evidence, mark it as NOISE instead of applying an unnecessary fix.

What should a code review finding include?

Each actionable finding needs a severity level, category, file:line evidence, concrete failure impact, the smallest safe fix, and an exact test or probe that verifies the fix. Style trivia handled by formatters is excluded.

What is the difference between code review and task verification?

Code review assesses implementation quality and spec fit of the diff, while a task verifier independently decides whether the task is complete and production-safe. The review does not duplicate the verifier's full release gate.

When should merge-base be used as the review comparison point?

Use the PR merge-base when reviewing a pull request and the fixed point is ambiguous. Otherwise state explicitly which comparison point was chosen, such as main or a user-supplied SHA.