skills · from skills/*/SKILL.md at v0.6.1

Skills

The playbooks an agent loads when the task matches. They come with the engine plugin; a pack adds its own beside them.

Skill	When
`act`	Do something about an entity in an Oto knowledge graph: list the actions the graph declares, confirm with a person before anything that changes the world, invoke the action through the transport it declares using your own tools, then record the response so it enters the graph as evidence through the curate gates. Use when someone says "what can I do about X", "create the repository", "check that it exists", "run the check", "call the API and record what it says", or asks to act on a fact the graph holds as intended. Oto never invokes anything; you do.
`build-knowledge-base`	Take a folder of documents all the way to a served, queryable knowledge graph with Oto: ingest them to Markdown, survey and read them, design the vocabulary from what they say, author the graph document by document as proposal files, merge and check them through the curate gates, build every layer, verify with real queries, and serve the result over JSON-RPC. Use when a user says "build a knowledge base from these documents", "turn this folder into a knowledge graph", or an Oto project exists but has no vocabulary or no graph yet.
`capture`	Turn knowledge that arrived in conversation into a dated, attributed source document in an Oto project's inbox, so it enters the graph through the same ingest, drafting, gates and pull request as any file. Use when someone corrects a fact, states a decision, or tells you something the graph should hold ("actually the owner changed in March", "remember that", "record this"), and especially when the project lives in a repository or the person is not a curator.
`concierge`	The guide to an Oto project: where it is in its life, what the choices are at this point, why Oto does what it does, and which playbook to load next. Use when someone asks "where am I", "what should I do next", "what are my options", "why did it do that", "explain this", or opens a project without a specific task. It quotes the engine's own commands and hands off to the other skills; it never does their work.
`curate`	Turn a document, or a statement someone made, into curated graph facts with provenance and dates. Four gates: recap what the source says, check contradictions against what is already believed, preview the exact diff, apply only on confirmation. Contradictions resolve by supersession, never by overwriting. Use whenever new knowledge must enter an Oto project.
`evaluate`	Validate an Oto knowledge graph against the documents it was built from: write a few gold questions per ingested document with the answer the document supports, run them against the graph, judge each answer for correctness and grounding, record the verdicts as feedback, and report what the graph gets right, wrong, and cannot answer. Use after a batch is authored and built, before `oto ingest complete`, or whenever someone asks "is the graph right?".
`ontology-interview`	Design a domain vocabulary for an Oto project by interview, and record why each class exists. Produces ontology.config.json and ontology.rationale.json, then validates both with the tooling. Use when a project has no vocabulary yet, or when an existing one needs restructuring.
`query-knowledge`	Answer a question from an Oto knowledge graph with temporal correctness: resolve the terms, read current facts only, cite the source and the date, frame what it means, and say plainly what the graph does not know. History and as-of dates only when asked. Read-only. Use whenever someone asks a question about the domain an Oto project covers, or wants an insight from it.
`spec`	Draft a product specification, an architecture decision record, or a design brief from what an Oto knowledge graph holds: the current facts as the ground, every claim cited to an entity and its source, open questions where the graph is silent, and the resulting decisions recorded back into the graph as first-class facts. Use whenever someone asks for a spec, an ADR, a design or the reasoning behind a choice, in a domain an Oto project covers.
`vet-provenance`	Audit an Oto knowledge graph so every fact traces to a document the corpus actually holds. Finds citations to documents that have left (de-scoped, renamed, replaced), separates a stale citation from a fact about to be lost, and produces a candidate graph that goes through the normal curate gates. Use when the document set has changed and you need to answer "is anything in here sourced to a file nobody can open?"