research-strategy-code

<skill_overview> <purpose>Optimize use of "assist/research/code-research" subagent for evidence-based codebase analysis</purpose> <triggers> <trigger>Locate where a feature or behavior is implemented</trigger> <trigger>Trace data flow or call chains across modules</trigger> <trigger>Explain configuration, environment, or initialization logic</trigger> <trigger>Answer questions with file/line evidence</trigger> <trigger>Inspect recent changes or regressions</trigger> </triggers> </skill_overview> <task_formulation> <principles> <principle>State the exact question and expected outcome</principle> <principle>Provide known entry points: file, symbol, or keyword</principle> <principle>Specify scope boundaries (module, package, layer)</principle> <principle>Request evidence with file paths and line numbers</principle> <principle>Specify depth level when needed: standard, deep, expert</principle> </principles> <examples> <good> <task>Where is JWT validation implemented? Show the handler path and middleware order with file references.</task> <why>Clear goal, scope, and evidence requirements</why> </good> <good> <task>Trace how user profile data is loaded from API to UI. Include data flow and key transformations.</task> <why>Call-chain focus and expected detail level</why> </good> <bad> <task>Explain the whole backend.</task> <why>Too broad and undefined scope</why> </bad> </examples> </task_formulation> <research_strategies> <strategy name="entry-point-discovery" use_for="Unknown start location"> <step order="1">Review project structure and likely entry files</step> <step order="2">Search for key terms or routes</step> <step order="3">Inspect candidate files to confirm entry points</step> </strategy> <strategy name="call-chain-trace" use_for="Behavior explanation"> <step order="1">Locate the handler or public API</step> <step order="2">Trace symbol usage across modules</step> <step order="3">Summarize the chain with evidence</step> </strategy> <strategy name="config-tracing" use_for="Runtime behavior"> <step order="1">Locate configuration definitions and defaults</step> <step order="2">Trace how configuration is loaded and applied</step> <step order="3">Document overrides and precedence</step> </strategy> <strategy name="change-context" use_for="Recent changes or regressions"> <step order="1">Identify recent changes and affected areas</step> <step order="2">Relate changes to observed behavior</step> <step order="3">Highlight potential regression risks</step> </strategy> <strategy name="multi-cycle" use_for="Complex or ambiguous questions"> <step order="1">Ask 1-3 targeted clarification questions</step> <step order="2">Research narrowly scoped answers</step> <step order="3">Repeat until scope is satisfied</step> </strategy> </research_strategies> <response_format>