compose:review

Dispatches a code reviewer subagent to evaluate git changes against requirements.

Updated Jun 27, 2026
One-click install
npx skills add https://github.com/HKUST-QUANT-SOCIETY/quantcode --skill compose-review-hkust-quant-society
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: compose:review
Source: https://github.com/HKUST-QUANT-SOCIETY/quantcode/tree/main/.opencode/meta-skills/review
Command: npx skills add https://github.com/HKUST-QUANT-SOCIETY/quantcode --skill compose-review-hkust-quant-society

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It catches defects, requirement gaps, and design issues in completed work before they compound across tasks or reach a merge, by routing changes through an independent reviewer with fresh context. ## Core Features & Use Cases - Subagent-Based Review: Dispatches a general subagent built from the code-reviewer.md template so the reviewer sees only the work product, not the session history. - Commit-Range Scoping: Uses BASE_SHA and HEAD_SHA to bound exactly which changes are evaluated. - Severity-Triaged Feedback: Classifies findings as Critical, Important, or Minor with clear rules for when to fix, defer, or push back. - Use Case: After finishing a task in subagent-driven development, request a review of the task's commit range, fix any Critical or Important issues, then proceed to the next task with confidence. ## Quick Start Ask the agent to request a code review of the changes between the previous commit and HEAD using the code-reviewer template before continuing to the next task.

Frequently Asked Questions about compose:review

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

FAQPage Schema
How do I request a code review for completed changes?

Get the base and head commit SHAs with git rev-parse, then dispatch a general subagent using the code-reviewer.md template filled with a description, the plan or requirements, and both SHAs. The reviewer returns strengths, issues by severity, and an assessment.

When should code review be requested during development?

Review is mandatory after each task in subagent-driven development, after completing a major feature, and before merging to main. It is also valuable when stuck, before refactoring, or after fixing a complex bug.

Why use a subagent for code review instead of self-review?

A subagent receives only the crafted review context, not the session history, so it evaluates the work product without bias from the implementation thought process. It also preserves the main session's context for continued work.

What should I do when the code reviewer is wrong?

Push back with technical reasoning rather than accepting incorrect feedback. Show code or tests that prove the implementation works, and request clarification on the disputed point.

What are the limitations of subagent-based code review?

The reviewer only sees the commit range and context provided in the prompt, so incomplete descriptions or wrong SHAs weaken the review. It also cannot replace human judgment for architectural or product-level decisions.