filter-building

To convert natural language intent into typed filters for code search, this skill maps vague requests into precise typed parameters like author, modifiedAfter, and language to refine semantic retrieval results.

Intent-aware filtering improves RAG search results by excluding noise from vendor directories, test files, and legacy code, ensuring the coding agent only references stable production implementations during bug fixes or domain-specific tasks.

Yes, you can filter code search results by git history metadata. The skill applies git-based metadata filters, allowing you to scope retrieval by specific authors, modifiedAfter dates, and commit counts to surface battle-tested code segments.

The best way to scope code retrieval and exclude test files or legacy directories is by combining file-level metadata filters, such as age and commit count, with chunk-level signals like test status and path patterns to isolate stable production code.

Yes, the TeaRAGs MCP server supports granular scoping for domain-specific code retrieval. It accepts structured typed filters combining path patterns, git-based metadata, and codegraph signals to restrict search results to relevant code segments.

Semantic code search results return unstable hotspots because they lack granular filtering. Applying typed filters for chunk-level signals like test status and symbol ID helps exclude noisy hotspots in favor of stable, specific code segments.