memory-entity-bootstraplisted
Install: claude install-skill ayushmall/memoryvault-kit
# memory-entity-bootstrap — pull source data about ONE entity, enrich it
## You are a sub-agent
You were spawned by `/memory-refresh` Step 4c (or by memory-master-ingest when creating a new entity for the first time). **You inherit the parent's MCP wholesale** — vault MCP + every source MCP the user has connected. Use them in parallel to pull what the kit knows about this entity across all sources.
Full contract: [`../../docs/AGENTS.md`](../../docs/AGENTS.md). The non-negotiables:
- **Before writing the entity file:** call `memory_search_entity(name=entity_name)` first. If it resolves to an existing canonical, enrich; don't duplicate.
- **For every memory you write:** check `source_ref` collision via `memory_search_entity` results before `memory_save`.
- **Read `.mvkit/learned_preferences.json` if present** — respect skip_authors / skip_channels.
- **Report back** in the structured shape: sources searched / items found / items already in vault / new memories written / entity file (created|enriched|unchanged) / skipped.
---
This skill is the deep-dive variant of ingest scoped to a specific
entity. The triggering case:
> memory-refresh just saw a Slack thread mentioning `[[Acme Corp]]`, but
> our vault only has 2 memories about Acme. Before saving the new
> memory and linking it, let's go pull what we know about Acme from
> all our source MCPs so the entity isn't a stub when the user asks
> about it later.
The kit's vault grows richer when entities are anchored properly