Read the Catalog Record
Read the Catalog Record
The catalog here is a generated file, catalog.json, rebuilt from the artifacts themselves whenever the library changes. Collections do not list their members; each artifact declares which collections it belongs to, and membership is derived. Delete an artifact and it leaves its collections automatically — a deliberate design choice, and a familiar one to anyone who has maintained a bib record and a holdings record separately.
When you ask "what's in this collection", something is actually being read: the plugin gave your assistant three catalogue tools, and they open the real index. That is worth noticing, because for most of this workshop you have been training yourself to assume the opposite. It is also worth testing — ask a follow-up about a detail and see whether the answer stays consistent.
The library-ai-workshop collection is the four Skills from this repository, imported into someone else's library, given a version, assigned a trust lane, and stamped with the commit they were taken from. This is your work as a catalog record. Read it the way you would read a vendor's description of your own institutional repository.
Survey the collections
The plugin gives your assistant three catalog tools: one lists collections, one lists artifacts, one reads an artifact in full. Something is genuinely being read rather than recalled. Start at the top.
List the Context Bazaar collections with their member counts and descriptions. Then tell me which ones an academic library might have any use for, and which are aimed at software teams.
You see six collections — kiro-official, neon-caravan, jh-drcc, byron-powers, library-ai-workshop, archon — with counts.