akb-querylisted

# AKB Query Answer **one question** from the target AKB vault. Decompose the question, search the vault with the strategy that fits, ground the answer in the documents found — honoring AKB's compiled-truth-over-timeline precedence — and synthesize a cited answer. Brain-first: the vault is the source of truth, not the model's prior knowledge. **Workflow classification.** LLM-wiki **Query** workflow — the **read** half only (à la gbrain's `query`, `mutating: false`). This skill **never writes**: no `akb_put` / `akb_update`, no promote-to-page. Promoting valuable answers into new pages, consolidation, and dream-cycle maintenance are the external maintenance layer's job over the same vault, not this skill's. Scope boundaries: - **Read-only.** Every tool call reads. The skill answers from vault content and cites it; it does not modify the vault. (It deliberately does **not** use `akb_grep` — that verb's `replace=` mode rewrites matching documents, so it is not read-only; exact-term lookup goes through `akb_search`, whose hybrid index already includes a keyword/BM25 component.) - **Brain-first, grounded.** Answer from the vault, never from general knowledge when the vault has relevant content. If the vault lacks the answer, say so — do not fill the gap with a guess. - **Citations are document-scoped.** AKB exposes provenance at the document level, not per claim, so a claim cites `[doc-uri, section]` — not a sentence-level source. ## Arguments ```text /akb-query {question} --v