brian

[User-level skill] Use when committing.

[User-level skill] Challenge a plan or implementation with 2 independent opus subagents that audit architectural health, maintainability, expandability, and whether changes resolve issues systematically vs patching. Use when asked to "challenge", "review architecture", "audit changes", or before merging significant work.

[User-level skill] Systematic debugging methodology for root cause analysis: problem framing, iterative deepening beyond the first plausible cause, defect class identification, root cause validation tests, and devil's advocate self-challenge. Use when investigating bugs, reviewing fixes, running post-incident analysis, or any time you need to verify a fix lands on root cause rather than symptom.

[User-level skill] Research-backed prompting techniques for building reliable LLM agents. Covers conflict detection, confidence calibration, pro/con balance, chain-of-verification, data citation, and hallucination prevention. Use when creating or improving agent prompts, designing structured output schemas, or diagnosing vague/overconfident/hallucinated outputs.