vestige-sources

name

vestige-sources

description

Use this skill when you need the raw source receipts for a memory or candidate — the typed evidence rows, not the narrative walk. Fire when the user says "what sources back mem_X?", "show me the source receipts", "list the sources for mem_X", or "filter sources by kind agent_session". Use `vestige-why` when you want the full provenance narrative including events; use this skill when you want the tabular source index only.

List source receipts

Return the typed source rows attached to a memory or candidate. Every memory captured with --source or promoted from a candidate carries at least one receipt. This skill gives you the raw evidence index — kind, reference, and optional content snippet — without the event narrative.

When to fire

• Evidence check. Confirming what files, commits, or sessions back a memory before citing it.
• Kind filtering. Narrowing to a specific source type — e.g. only agent_session sources, or only file sources.
• Pre-approval audit. Inspecting a candidate's sources before deciding whether to approve or reject.
• Bulk source review. A memory has many sources and you want the table, not the prose.

How to invoke

vestige sources <mem_or_cand_id> --json

• --kind <kind>: filter to one source kind. Valid kinds: file, commit, url, agent_session, mcp_call, candidate, manual, trace.
• Accepts mem_<ULID> or cand_<ULID>.

vestige sources mem_01JWXXXXXXXXXXXXXXXXXX                       # all sources, text
vestige sources mem_01JWXXXXXXXXXXXXXXXXXX --kind agent_session  # filtered
vestige sources cand_01JVXXXXXXXXXXXXXXXXXX                      # candidate sources
vestige sources mem_01JWXXXXXXXXXXXXXXXXXX --json                # JSON output

After invocation

{
  "owner_id": "mem_01JWXXXXXXXXXXXXXXXXXX",
  "sources": [
    {
      "id":         "src_01JWXXXXXXXXXXXXXXXXXX",
      "kind":       "candidate",
      "source_ref": "cand_01JVXXXXXXXXXXXXXXXXXX",
      "content":    null,
      "truncated":  false
    },
    {
      "id":         "src_01JVXXXXXXXXXXXXXXXXXX",
      "kind":       "agent_session",
      "source_ref": "current",
      "content":    "Use ULID over UUID for all entity IDs — lexicographic sort…",
      "truncated":  false
    }
  ]
}

Source kinds: file · commit · url · agent_session · mcp_call · candidate · manual · trace. Use vestige-why <id> for the full event walk alongside sources.