| name | vestige-trace-show |
| description | Use this skill when you want to expand a specific trace to its full detail — the query text, mode requested vs resolved, provider, result IDs, and scores. Fire when the user says "expand trace_X", "show me what trace_X returned", or "what did that search find?". Use vestige-trace-list first to find the trace_id, then this skill to inspect one in full. |
Expand a single query trace
Return the full detail for one query_events row: kind, mode (requested vs resolved), provider, parameters, result IDs with scores, and score methodology. Use this after vestige-trace-list to drill into a specific trace.
When to fire
- Inspect a recall result set. "What did that hybrid search find?" → expand the trace to see IDs and scores.
- Debug mode resolution. Confirm whether
hybrid was requested and resolved, or fell back to lexical.
- Provider audit. Check which embedding model was in use when a trace was written.
- Score breakdown. Understand how results were ranked (RRF fusion, vector similarity, or FTS).
How to invoke
vestige trace <trace_id> [--json]
- Accepts a
trace_<ULID> — obtain one from vestige trace (list mode) or the trace_id field in any trace envelope.
vestige trace trace_01JWXXXXXXXXXXXXXXXXXX
vestige trace trace_01JWXXXXXXXXXXXXXXXXXX --json
After invocation
{
"trace_id": "trace_01JWXXXXXXXXXXXXXXXXXX",
"kind": "search",
"mode_requested": "hybrid",
"mode_resolved": "hybrid",
"caller": "mcp",
"query": "ULID migration ordering",
"provider": "fake",
"provider_model": "deterministic-sha256",
"params": {
"limit": 8,
"type_filter": null
},
"result_count": 2,
"result_ids": ["mem_01JWXXXXXXXXXXXXXXXXXX", "mem_01HVXXXXXXXXXXXXXXXXXX"],
"result_scores": [0.83, 0.61],
"latency_ms": 43,
"created_at": "2026-05-08T14:02:11Z"
}
To re-run a trace against the current store and diff the results, use vestige-trace-replay <trace_id>.
