ozerohax/task-use-research-code-strategy

<when_to_use> <trigger>Need to understand architecture or data flow across multiple files</trigger> <trigger>Find where a feature is implemented and how it works end-to-end</trigger> <trigger>Locate patterns or examples across the codebase</trigger> <trigger>Need a map of modules, dependencies, or ownership boundaries</trigger> </when_to_use>

<task_request> <principles> <principle>State the goal and expected output (summary, file list, call chain)</principle> <principle>Include starting points if known (file paths, symbols, errors)</principle> <principle>Limit scope: 1 topic or feature per request</principle> <principle>Specify depth level when needed: standard, deep, expert</principle> <principle>Ask for evidence: require references to files and functions</principle> </principles> <examples> <good> <task>Explain how auth works end-to-end. Provide key files, entrypoints, and data flow.</task> <why>Clear goal, expects evidence and structure</why> </good> <good> <task>Find all places where Redis is used and summarize why.</task> <why>Specific target and output</why> </good> <bad> <task>Review the whole codebase</task> <why>Too broad, no specific objective</why> </bad> </examples> </task_request>

<research_principles> <principle>Start with entrypoints (main, routes, handlers, controllers)</principle> <principle>Trace data flow across layers: API -> service -> persistence</principle> <principle>Prefer reading source to inference; avoid assumptions</principle> <principle>Collect exact file paths and symbol names as evidence</principle> <principle>Separate confirmed facts from hypotheses</principle> <principle>Note configuration and environment dependencies that affect behavior</principle> </research_principles>

<research_strategies> <strategy name="map-first" use_for="Unknown codebases"> <step order="1">Identify top-level structure and entrypoints</step> <step order="2">Find key modules and ownership boundaries</step> <step order="3">Dive into the smallest set of files that answer the question</step> </strategy> <strategy name="trace-flow" use_for="End-to-end behavior"> <step order="1">Start from the trigger (API, event, CLI)</step> <step order="2">Follow the call chain across layers</step> <step order="3">Document data transformations and side effects</step> </strategy> <strategy name="pattern-hunt" use_for="Examples and reuse"> <step order="1">Search for similar modules or keywords</step> <step order="2">Compare implementations and extract common patterns</step> <step order="3">List best candidates with file references</step> </strategy> </research_strategies>

<output_requirements> <requirement>Provide a short summary of findings</requirement> <requirement>List key files with paths and roles</requirement> <requirement>Highlight open questions or missing context</requirement> </output_requirements>

<agent_limitations> <cannot>Edit or write files</cannot> <cannot>Run build or tests</cannot> <cannot>Remember previous sessions</cannot> </agent_limitations>

<depth_levels> <level name="standard">High-level flow + key files</level> <level name="deep">Call chain, data transformations, edge cases</level> <level name="expert">Design tradeoffs, risks, and refactor targets</level> </depth_levels>