in-the-loop-labsOfficial·10 Agent Skills Included

pair-review

Human-in-the-loop code review with multi-model analysis

Reviews GitHub pull requests and local changes in a familiar diff UI with three-level analysis across changed lines, file context, and codebase architecture. Runs multi-model council reviews, collects inline suggestions, and feeds structured feedback back to coding agents for iterative fixes. Eliminates manual diff reading and keeps humans in control of every review decision.

48|18

↗

npx skills add in-the-loop-labs/pair-review --all -g -y

Available:

Gives your AI agent the project requirements, review workflow rules, and coding constraints it must follow when building or modifying this code review application.

All Skills in This Repository (10)

Pure Emerald Level Indicators

📦 In Repo

in-the-loop-labs

review-requests

Open pending GitHub pull requests in the pair-review web UI with auto-analysis.

ai-critic

Fetches and applies AI-generated code review suggestions from pair-review.

local

Open uncommitted local code changes in the pair-review web UI.

pr

Open the current branch's GitHub pull request in the pair-review web UI.

user-critic

Fetch pair-review comments and apply code changes in local or PR contexts.

analyze

Analyze code via the pair-review MCP server with configurable tiers.

Update Provider Models

Update AI provider model configurations in pair-review source code files.

loop

Orchestrate iterative code implementation, AI analysis, and issue fixing cycles.

review-roulette

Dispatch code review tasks to three random reasoning models and merge attributed summaries.

review-model-guidance

Guide AI model selection for code review tasks by complexity and cost.

Official

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install pair-review?▼

Run `npx skills add in-the-loop-labs/pair-review --all -g -y` in your terminal to install all skills in this suite globally.

How does pair-review help with code reviews?▼

It opens pull requests or local changes in a GitHub-like web UI and runs three levels of analysis, from changed lines up to full codebase architecture. You adopt, edit, or dismiss each suggestion before submitting your review.

What is the pair review loop skill?▼

The loop skill runs repeated council reviews of your code, fixes the findings, and re-reviews until a final round reports no merge blockers. It keeps the human informed with a verdict report after every run.

Which AI providers does pair-review support?▼

It works with Claude, Codex, Copilot, Cursor, OpenCode, Antigravity, Pi, and Muse through their command-line tools. You can use your existing subscriptions with no extra cloud service.

Does pair-review send my code to the cloud?▼

No. It is local-first, so all review data and processing stay on your own machine in a local SQLite database.