| name | vault-query |
| description | Search and analyze the vault knowledge base. Query learnings by type, status, tags, or pattern-key. Show promotion candidates and recurring patterns. |
| user-invocable | true |
| argument-hint | <query | stats | promotions | recent | by-tag <tag>> |
| context | fork |
| agent | researcher |
| allowed-tools | ["Read","Glob","Grep"] |
| model | sonnet |
Vault Query
Search and analyze the vault knowledge base from the CLI.
Subcommands
Parse $ARGUMENTS to determine which subcommand to run:
stats
Count learnings by status and type, show promotion candidates:
Glob for vault/learnings/*.md- Read each file's frontmatter
- Aggregate counts by
type (learning/error/feature), status, area, and project
- List any entries with
recurrence-count >= 3 as promotion candidates
- Output a summary table
promotions
List learnings eligible for promotion:
Grep for recurrence-count: in vault/learnings/*.md- Read files where
recurrence-count >= 3
- For each, show: id, title, recurrence-count, first-seen, last-seen, pattern-key
- Sort by recurrence-count descending
recent (or recent N)
Show the last N learnings (default 10):
Glob for vault/learnings/*.md- Sort by filename (date-based naming gives chronological order)
- Read the last N files
- Show: id, type, title, status, tags
by-tag <tag>
Filter learnings by tag:
Grep for the tag in vault/learnings/*.md frontmatter tags: lines- Read matching files
- Show: id, type, title, status
Free-form query
For any other $ARGUMENTS, treat it as a search query:
Grep for the query in vault/learnings/*.md (both frontmatter and body)- Also search
vault/shared/**/*.md
- Read matching files and present relevant excerpts
- Highlight matching
pattern-key, tags, and section content
Output Format
Use concise tables for structured output. Example:
| ID | Type | Title | Status | Tags |
|----|------|-------|--------|------|
| LRN-20250310-001 | learning | Claude CLI env stripping | resolved | [cli, env, spawn] |
For stats, use summary format:
Vault Stats:
Learnings: 12 (8 resolved, 3 new, 1 promoted)
Errors: 5 (4 resolved, 1 investigating)
Features: 3 (1 built, 2 requested)
Promotion candidates: 2
