What problem does it solve? Maintaining a knowledge base manually fails because bookkeeping — preserving sources, cross-linking pages, tracking superseded notes — costs more effort than the value it returns. This Skill encodes the layer model, folder rules, status flows, and log discipline so an agent can run the ingest → research → consolidate loop without losing the evidence chain. ## Core Features & Use Cases - Three-layer layout: Governs external-sources/ (verbatim raw sources), research/ (provisional draft analysis), and articles/ (canonical stable knowledge) with per-folder rules. - Status and supersedes flows: Tracks status: draft → stable transitions and maintains supersedes: / superseded_by: chains so every claim traces back to a preserved source. - Log discipline: Requires one dated log.md entry per content-changing turn, with markdown links so entries register in the knowledge graph's backlinks. - Use Case: A user shares a URL; the agent ingests it verbatim into external-sources/, drafts a sourced research note, and after a team decision consolidates a canonical article that supersedes the research — all logged and linked. ## Quick Start Ask the agent to ingest a shared URL into the knowledge base and then research the topic using the preserved sources.
