second-brain-query
How to query second-brain-librarian efficiently — ask for exactly what you need, not "everything about X." Read before invoking it in query mode (coder checking for a reusable pattern, project-bootstr
SKILL.mduser-level-skills/second-brain-query/SKILL.md
GitHubQuerying the second brain efficiently
second-brain-librarian reads the vault's own CLAUDE.md and can produce a thorough answer —
but a vague, broad question makes it read more of the vault and write a longer answer than the
task needs. Ask narrowly.
Rules
- Ask for the specific thing, not the topic. "What's the current Verse snippet and Visto/testato su list for the Elimination Manager pattern?" — not "tell me everything about Elimination Manager."
- Prefer a
type: synthesisarticle when one might exist. If you're about to ask about two or three related concepts together, ask first whether a synthesis article already combines them — one citation instead of stitching together several component articles yourself. - State what you'll do with the answer, briefly — it helps the librarian judge how much detail actually matters (a snippet to adapt needs the code; a track-record check just needs the outcome, not the full article).
- Don't re-query for the same thing twice in one session — if you already asked and got "no match," don't ask again a different way hoping for a different answer; proceed as if the pattern doesn't exist yet.
- Never ask it to summarize the whole vault or a whole thematic wiki "just in case" — that's exactly the kind of broad request this skill exists to avoid.
Names and free checks
The vault may mix English and Italian names. names.json (next to this file) maps each canonical English name to the Italian ones; scripts/vault_check.py names|links|stats checks the vault at no AI cost.