memory-deep-divelisted

For a thin-retrieval queue item, fetch richer content from the source's native MCP, synthesize into a properly-shaped memory, save it. Use when invoked by /memory-refresh Step 4b for "deep-dive" classified items — or when the user explicitly says "deep-dive this query" / "fetch X from Notion" / "go to source for Y". Specialized per source (Notion, Slack, Linear, Gmail, GDrive, Granola, GitHub).
ayushmall/memoryvault-kit · ★ 1 · AI & Automation · score 75

Install: claude install-skill ayushmall/memoryvault-kit

# memory-deep-dive — escape to the source ## You are a sub-agent You were spawned by `/memory-refresh` (or directly invoked for one query). **You inherit the parent session's MCP wholesale** — the vault MCP (`mcp__plugin_memoryvault-kit_memoryvault__*`) AND every source MCP the user has connected (Slack, Linear, Notion, Gmail, Granola, GitHub, GDrive, Pylon, …). Use them; don't read the vault from disk. The full handover contract is [`../../docs/AGENTS.md`](../../docs/AGENTS.md). The bits you must honor every run: - **Before saving:** dedupe with `memory_search_entity` (entities) + `source_ref` collision check (memories). - **Read `.mvkit/learned_preferences.json` if present** — respect `source_overrides.<source>.skip_*` and `filter_rules.*`. - **Report back** in the structured shape from AGENTS.md §4 (sources searched / items found / new memories written / skipped / notes). --- **One job**: when the vault's existing memories don't answer a query, go to the source's native MCP and pull richer content. Synthesize that into a new memory the kit will retrieve next time. ## The input You receive (from `memory-queue-router` or the user directly): - A `query` (the original question that came back thin) - A `suggested_source` (which native MCP to query — inferred from the partial results' `parent_surface:` field) - The partial `result_ids` (what the kit DID find, however weakly) ## The flow ### 1. Translate the query The original query is in the user's words. Translate