Multi-Specialist Review

Team-based code review that spawns 3-5 parallel specialist sub-agents to examine a diff through different perspectives (correctness, security, performance, architecture, etc.), mechanically verifies every finding against the actual source files, and synthesizes a single prioritized review artifact.

When to Use

This skill is user-triggered. It is not part of routine agent workflows — the cost and authority profile warrants explicit user invocation. Use when:

• PR-level changes (5+ modified files)
• Multi-commit ranges (main..feature-branch) rather than atomic commits
• Security-sensitive paths (auth, crypto, payments, input validation)
• Single-turn specialist-review flagged quality concerns that need deeper scrutiny

For atomic-commit review, use the baseline specialist-review.sh from the agent-loops skill instead — it's 4-6x cheaper and sufficient for small diffs.