Prompt (rr-iter2-004): Refactor `skill_advisor.py` to surface raw ambiguity counts in debug output.
Expected Gate 3: true

Classifier output:
{
  "triggersGate3": true,
  "reason": "file_write_match",
  "matched": [
    {
      "pattern": "refactor",
      "kind": "token",
      "category": "file_write"
    }
  ],
  "readOnlyMatched": []
}