Use for strict code-quality review of local changes, commits/ranges, or GitHub PRs; also for review-oriented subagents/checkpoints in orchestrated coding work.
Use to delegate context-heavy or independent work to an isolated worker: broad code search, web research, multi-file exploration, long-running implementation, testing/log investigation, review, parallel analysis, or concise summarization of noisy work.
Use when compressing context for another session, agent, fork, subagent, queued item, or future continuation. Defines handoff content standards, not file/session mechanics.
Use when coordinating a long-running coding effort through plans, subagents, handoffs, reviews, queues, and checkpoints while keeping the main context small.
Test-driven development with red-green-refactor loop. Use by default when implementing features, fixing behavior bugs, changing public behavior, or adding integration tests; also use when user mentions testing, TDD, red-green-refactor, or test-first development.
Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or says "grill me on..." or similar.
Research current, external, or web-based information; compare products, pricing, libraries, APIs, companies, docs, releases, issues, or news; and analyze content from URLs.