rag-your-code

RAG Your Code

Use this skill when repository context is broad or a symbol's implementation is difficult to locate.

Four commands cover the same ground explicitly, and a user can reach them without this skill firing at all: /rag-your-code:index, /rag-your-code:search, /rag-your-code:describe, /rag-your-code:status. Offer the one that fits rather than narrating the steps, and offer /rag-your-code:describe whenever coverage is the thing holding retrieval back — it is the largest available improvement and the one nobody discovers on their own.

Workflow

1. Make sure the package is importable. A plugin install copies this skill but not the Python package, so check once per machine and install if the import fails. Use python -m pip rather than a bare pip so the package lands in the same interpreter the later commands run under.

   python -c "import ragyourcode" || python -m pip install --user rag-your-code
   

2. Build the index and find out what this repository still needs:

   python -m ragyourcode.cli bootstrap .
   

   Indexing a repository is not the same as making it searchable. A fresh index retrieves against the sentence the parser generated, which adds no word the source did not already have, so a question phrased in domain terms reaches nothing. bootstrap indexes, says which rung this repository is on — descriptions to write, a promotion to apply, or nothing left — and hands over that rung's work. It reads the state rather than remembering a position, so running it again after each round is how you make progress.

   index still exists and does only the indexing.

3. Retrieve focused context before making a change:

   python -m ragyourcode.cli search "<concept or behavior>" --json --limit 8
   

   Add --graph --hops 1 when callers, callees, containment, or small imported modules are relevant. Prefer one hop; use two only when the first path provides concrete edge evidence.