Навык codebase-analyzer

Codebase Analysis

Inspect the portion of the repository needed for the question. Start with the working tree, manifests, entry points, and nearby tests. Use an existing code-to- document map when it helps locate relevant evidence, then verify claims in code.

Evidence to collect

• Languages, frameworks, package managers, runtime versions, and build commands.
• Entry points, routes, services, data stores, and external integrations.
• Data and control flow through the affected feature, including errors.
• Naming, module boundaries, dependency patterns, and test conventions.
• Configuration, migrations, CI checks, deployment files, and operational hooks relevant to the requested change.

Follow a representative request or operation end to end. Link findings to actual files and symbols. Distinguish declared dependencies from runtime use and generated code from maintained source. Scale exploration to uncertainty.

For a project profile, summarize stack, directory responsibilities, key flows, tests, risks, and the commands needed to reproduce findings. Build a feature inventory from reachable code or documented behavior, identifying evidence gaps. For implementation work, use the findings directly to choose and verify the change. A standalone analysis request ends with the requested explanation.