learnings-researcher

To search an internal knowledge base for prior solutions, you can run a grep-first discovery process against project learnings and an optional INDEX.md, parsing frontmatter to distill the top entries into root causes and actionable recommendations.

Finding recorded institutional knowledge about a production timeout involves querying the knowledge-base directory, where grep pre-filters candidates and frontmatter metadata prioritizes results by severity and tag matching to surface tested fixes.

You can use frontmatter metadata to prioritize internal learnings for debugging by matching module tags and severity levels, allowing the system to deep-read the top 5-10 files and extract the root cause and solution patterns.

Index-first discovery works by checking an optional knowledge-base INDEX.md for broad domain matches to find high-level pointers, which then narrows down candidates using targeted case-insensitive grep patterns before deep reading the files.

A limitation of relying on grep-based pre-filtering for a project knowledge base is that it requires targeted case-insensitive patterns to effectively narrow candidates, meaning unstructured or poorly tagged entries may be missed during discovery.

You do not need an INDEX.md file to surface internal learnings for integration tasks, as the system operates on project learnings directly using grep-based candidate discovery, although the index is preferred for broad domain matches.