Skill Instructions

Inputs / Outputs / Non-goals

• Inputs: user questions, relevant PRDs/docs, RAG queries and results.
• Outputs: investigation summary, sources, unknowns, recommended next steps.
• Non-goals: implementation changes without a follow-up task.

Trigger conditions

Use this skill when prompts include or imply:

• Investigating bugs or unclear behavior.
• Planning new features before coding.
• Answering architecture/workflow questions.

Mandatory rules

• Query RAG first for discovery; retry once with a rephrased query if weak.
• Summarize findings with provenance from docs/code.
• Identify unknowns and assumptions explicitly.
• Fallback to targeted file search when RAG is insufficient.

Validation checklist

• At least one RAG query executed for the investigation topic.
• Sources are cited from PRDs/docs/code.
• Unknowns and next steps are explicit.
• Fallback search performed when RAG quality is weak.

Expected output format

• Investigation summary.
• Source list (RAG and/or direct file inspection).
• Recommended next actions and open questions.

Failure/stop conditions

• Stop after two weak RAG attempts and report fallback findings + remaining gaps.
• Stop when missing context blocks reliable recommendations.