dev-review

Runs dual engineering and QA review passes over a change scope and synthesizes findings into analysis.md.

154|57|Updated Feb 7, 2020
One-click install
npx skills add https://github.com/FHIR/fhir-codegen --skill dev-review-fhir
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: dev-review
Source: https://github.com/FHIR/fhir-codegen/tree/main/.github/skills/dev-review
Command: npx skills add https://github.com/FHIR/fhir-codegen --skill dev-review-fhir

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Pre-PR code changes often ship with undetected antipatterns, coverage gaps, and regression risks because a single review pass misses half the picture. This Skill runs two independent staff-level reviews — one Engineering Lead pass and one QA Lead pass — over a defined change scope, then synthesizes both critiques into a single ranked, actionable analysis.md report. ## Core Features & Use Cases - Two-track independent review: An Engineering Lead pass covers antipatterns, hot paths, consistency with AGENTS.md conventions, dead code, and design issues; a QA Lead pass covers test coverage, edge cases, regression risk, and verifiability. The passes never see each other's output until synthesis. - Flexible scope resolution: Review the working tree, last commit, commits since push, a commit range, a file list, or the commits recorded in a sibling plan.md's progress log. - Synthesized severity-ranked report: Findings are deduplicated, ranked Blocker through Nit, cited with file paths and line ranges, and each ends with a concrete recommendation. - Use Case: After dev-do produces a few commits in scratch/0423-02/, run the review with slot number 2 to get a read-only quality gate report before opening the PR. ## Quick Start Review the commits from slot 2 and write the synthesized engineering and QA findings to its analysis.md file.

Frequently Asked Questions about dev-review

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

FAQPage Schema
How do I run a code review on my local commits before opening a PR?▼

Invoke the review with a target slot number or output path and a scope such as since-push, last-commit, or working-tree. It runs independent Engineering and QA passes over the resolved commits and writes a severity-ranked analysis.md with cited findings.

What does the QA review pass check in a change set?▼

The QA pass checks test coverage of new code paths, edge cases like empty inputs and concurrency, regression risk against existing behavior, test determinism and flakiness, assertion quality, and whether logs and metrics support debugging failures in production.

Can the review scope be derived from an existing plan file?▼

Yes. When the analysis path resolves to a slot directory containing a plan.md, the review reads its Progress Log, collects every COMMIT entry's SHA, and reviews exactly that commit set, ignoring PENDING and NOTE entries.

Does this review modify source code or commit anything?▼

No. The review is strictly read-only with respect to the codebase: it never edits source, stages, commits, or pushes. The only file it writes is the analysis.md report under the gitignored scratch directory.

How do I limit parallel sub-agents during a large review?▼

Pass the max_subagents parameter, which defaults to 3 and caps at 8. Setting it to 1 disables parallel fan-out so the Engineering and QA passes run sequentially while remaining independent of each other.