session-search
Keyword search across Claude Code session transcripts.
This skill should be used when searching Claude Code session transcripts with semantic understanding. Triggers on queries like "find sessions about X", "when did I work on Y", "search previous conversations". Supports natural language queries with synonym matching.
What it does
Searches Claude Code session transcripts by combining keyword pre-filtering with semantic evaluation. Faster than session-finder (no embedding index needed) but relies on keyword matching plus LLM-based relevance scoring.
How it works
- Keyword pre-filter — scans all sessions for keyword matches against your query
- Excerpt extraction — pulls meaningful excerpts from top candidate sessions
- Semantic evaluation ��� scores each session’s relevance considering synonyms (“bug” matches “error”), related concepts (“debugging” matches test failures), and domain context
Key features
- Synonym matching — “refactoring” matches Edit-heavy sessions; “obsidian” matches vault-related work
- Configurable — adjust max results and lookback period (default: 10 results, 90 days)
- No indexing required — works immediately without building an embedding index first
When to use
When you need to find a past session quickly without waiting for an index build. For more precise semantic search over large session histories, use session-finder instead.