skills.rest
visdomtech
visdomtechOFFICIAL
Explore all skills by this author →
WebsiteGitHub

multi-agent-review

Dispatches seven parallel review agents to audit code changes with build verification and readiness scoring.

Updated Apr 28, 2026
Search Tags:#static analysis#code review#pull request#github#security audit#multi-agent#dev-cycle
One-click install
npx skills add https://github.com/visdomtech/skills --skill multi-agent-review-visdomtech
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: multi-agent-review
Source: https://github.com/visdomtech/skills/tree/main/multi-agent-review
Command: npx skills add https://github.com/visdomtech/skills --skill multi-agent-review-visdomtech

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Single-pass code reviews miss cross-cutting concerns like security flaws, race conditions, and lifecycle bugs, especially in high-stakes changes such as auth, payments, migrations, and large refactors. ## Core Features & Use Cases - 7-Agent Parallel Review: Dispatches specialized agents (Architect, Logician, Sentinel, Performer, Cartographer, Observer, Tracer) covering architecture, correctness, security, performance, completeness, observability, and lifecycle transitions. - Mandatory Pre-Flight Verification: Runs build, test, lint, and typecheck before review so agents consume verified status instead of guessing. - Structured Verdict with Auto-Fix: Produces a severity-tagged report with a 0-100 readiness score, posts it to GitHub in PR Mode, and hands findings to dev-cycle for automated fixing and re-review. - Use Case: Before merging a 20-file refactor touching session handling and database migrations, invoke an ultra review to catch stale-reference bugs, N+1 queries, and schema drift that a standard review would miss. ## Quick Start Ask the AI to run multi-agent-review on PR #42 or on your current uncommitted changes for an ultra review.

Frequently Asked Questions about multi-agent-review

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

FAQPage Schema
How do I run a multi-agent code review on a GitHub PR?▼

Invoke multi-agent-review with a PR number or URL, such as "review PR #42". The skill fetches the PR via the gh CLI, resets HEAD to the PR head, runs build and test verification, then dispatches seven review agents in parallel.

What is the difference between pr-review and multi-agent-review?▼

multi-agent-review is a self-contained superset of pr-review that uses 7 specialized agents instead of 3. Every pr-review lens maps to an agent, and it adds deeper dependency tracing, lifecycle analysis, and a numeric readiness score.

Can I review local commits or uncommitted changes without GitHub?▼

Yes, Local Mode reviews commit ranges or uncommitted WIP changes read-only, with no gh commands, pushes, or branch switching. Provide a COMMITS range or let the skill infer the target from your working tree.

How does incremental review work for already-reviewed PRs?▼

The skill detects its own prior review on the PR and reviews only the delta commits pushed since that review. Set the FULL flag or ask for a full review to re-examine the entire base-to-head range instead.

What happens after the review finds issues?▼

Any finding of any severity triggers Phase 7, which hands a structured fix brief to dev-cycle for resolution under a user-approval gate. PR Mode then pushes fixes and re-reviews; Local Mode commits locally and stops, with at most one automatic fix iteration.

When should I not use multi-agent review?▼

For small changes of 3 files and 50 lines or fewer, the skill automatically falls back to a single-pass review applying all 7 lenses without subagents. Routine low-risk changes are better served by the lighter pr-review skill.

Related Skills

Explore More Software Engineering Skills→
windmill-labswindmill-labs
17.5k

pr

Create draft GitHub pull requests with conventional commit titles.

Official
Advanced
az9713az9713
2

land-and-deploy

Merge pull requests, deploy code, and verify production health.

Community
Advanced
uzumaki-incuzumaki-inc

health

Run code health checks and compute a weighted 0-10 score.

Official
Advanced
razorpayrazorpay
31

risk-assessment

Classify pull requests in the agent-skills repository as LOW, MEDIUM, or HIGH risk.

Official
Intermediate