reference-lookup

name: reference-lookup description: Look up relevant research content to support coaching during quiz or teach-back sessions

Research Reference Lookup

Find and extract relevant research content for a given concept or question. Used during interactive coaching sessions to ground feedback in source material.

Input

• {{QUERY}} -- the concept, term, or question to look up
• {{RESEARCH_DIR}} -- path to the directory containing research markdown files

Procedure

1. Extract key terms. Parse {{QUERY}} to identify the most important search terms. Consider synonyms and related terms (e.g., "caching" might also appear as "cache", "cached", "memoization").

2. Search across research files. Use Grep to find mentions of the key terms across all markdown files in {{RESEARCH_DIR}}. Search broadly first, then narrow down.

3. Rank results by relevance. Prioritize files and sections where:

   • Multiple key terms appear in close proximity
   • The term appears in a heading (indicating a dedicated section)
   • The content provides explanations, not just passing mentions

4. Read the most relevant files. Read the top 1-3 files identified by the search. Do not read every file in the directory.

5. Extract focused excerpts. Pull out the specific paragraphs or sections that directly address {{QUERY}}. Do not return entire files. Each excerpt should be self-contained and useful for coaching.

6. Identify related topics. Note any other topics or concepts referenced in the excerpts that the learner might want to explore next.

Output Format

Return a single JSON object:

{
  "found": true,
  "excerpts": [
    {
      "source": "research/03-caching-strategies.md",
      "content": "The relevant paragraph or section text extracted from the file..."
    },
    {
      "source": "research/07-performance.md",
      "content": "Another relevant section from a different file..."
    }
  ],
  "relatedTopics": [
    "cache invalidation",
    "CDN configuration",
    "TTL strategies"
  ]
}

• found: boolean, whether any relevant content was located
• excerpts: array of source-attributed content snippets (aim for 1-3)
• relatedTopics: array of related concept names found in or near the excerpts

When Nothing Is Found

If no relevant research content is found, return:

{
  "found": false,
  "excerpts": [],
  "relatedTopics": []
}

Notes

• Keep excerpts concise. A few paragraphs per excerpt is ideal. Do not dump entire files.
• Preserve the original wording from the research files. Do not paraphrase or summarize.
• If the query is ambiguous, search for multiple interpretations and include the most relevant results.
• This skill is a support function called by other skills (evaluate-rationale, score-explanation). Speed matters.