analyze

name	analyze
description	Perform AI-powered code review analysis by spawning parallel Task agents directly within the coding agent's context. Does not require the pair-review MCP server — works standalone. Runs Level 1 (diff isolation), Level 2 (file context), and Level 3 (codebase context) as parallel tasks, then orchestrates results into curated suggestions. Results are returned directly in the conversation and also pushed to the pair-review web UI (if running). Use when the user says "analyze", "analyze my changes", "run analysis", "analyze using tasks", "analyze directly", "analyze here", or wants code review analysis of their changes. This is the default analysis skill. If the user says something ambiguous like "analyze my changes" or "run analysis", use this skill unless they specifically ask for in-app analysis.
arguments	{"tier":{"description":"Prompt tier: fast (surface/Haiku-class), balanced (standard/Sonnet-class, default), or thorough (deep/Opus-class)","required":false,"default":"balanced"},"skipLevel3":{"description":"Skip Level 3 (codebase context) analysis. Useful for small or isolated changes.","required":false,"default":false},"customInstructions":{"description":"Optional repo or user-specific review instructions (e.g., 'focus on security' or 'this repo uses X pattern')","required":false}}

Analyze Changes

Perform a three-level code review analysis and return curated suggestions.

Tools

This skill includes a scripts/git-diff-lines script that annotates git diff output with explicit OLD and NEW line numbers. Each subagent should invoke it by name (git-diff-lines) — the orchestrating agent must ensure the script's directory is on PATH (e.g., via PATH="<skill-dir>/scripts:$PATH").

1. Gather context

Determine what's being reviewed: