code-review-amplifier

Installation
SKILL.md

Code Review Amplifier

Core principle: A perfect code review serves seven concurrent functions on a shared artifact. AI can fully handle 2/7, partially handle 2/7, and barely touch 3/7. This skill maximizes total review quality by doing what AI does well (surface scanning, velocity) while arming the human reviewer with context and questions for the dimensions that require human judgment (design coherence, knowledge transfer, mentoring).

The goal is not to produce a review. The goal is to make the human reviewer's next 15 minutes dramatically more effective.


The Seven Dimensions of Code Review

Every code review, consciously or not, operates across these dimensions. Most reviews only cover 2-3 of them. A "perfect" review touches all seven with appropriate depth.

# Dimension Core Question AI Role
D1 Correctness Does this code do what it claims? Pre-scan: Flag logic issues, edge cases, type mismatches, missing error handling
D2 Design Coherence Does this fit the system's architecture? Arm the human: Surface relevant architectural context, generate design questions
D3 Readability Can the next person understand this? Pre-scan: Flag complexity, naming, structure, readability issues
D4 Security & Resilience Does this introduce vulnerabilities? Pre-scan: Check for common vulnerability patterns, data exposure, failure modes
D5 Knowledge Transfer Do more people now understand this area? Route: Suggest who else should see this code and why
Related skills

More from andurilcode/craftwork

Installs
2
GitHub Stars
6
First Seen
Apr 2, 2026