kramme:find-bugs:team

A team-based security review on large code diffs is executed by spawning four teammates in delegate mode. This orchestrates parallel analysis to cross-check injection, auth, data leakage, and logic issues across the shared diff.

Yes, you must enable parallel code analysis with agent teams by setting the CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS environment variable. This prerequisite allows the skill to spawn multiple teammates for cross-validation.

The best way to cross-validate security findings is mapping attack surfaces and enforcing a structured checklist across multiple reviewers. This team-based approach produces verifiable outputs by checking competing hypotheses against the diff.

You should use a multi-agent security review instead of a standard code review when targeting large diffs, security-critical components, or when competing hypotheses are needed. It maps attack surfaces and cross-checks findings across multiple teammates.

Yes, team-based code review works without external dependencies, but requires the CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS feature enabled. It spawns four teammates directly to execute the structured security checklist.