enhancement: literature-anchored improvements for the knowledge-engine & documentation roles (knowledge-*, documenter, technical-writer

> Generated from a literature-anchored content review of the generic agent roles (`agents/1-generic/`) against our internal literature index (books, section + similarity score) and verified online standards (URLs HTTP-checked; key claims confirmed by full-text/`web_extract`). > Each item = concrete improvement suggestion + target location + sources. Review scope: agent-meta v1.1.0. ## Scope `knowledge-curator`, `knowledge-ingestor`, `knowledge-querier`, `knowledge-linter`, `knowledge-indexer`, `knowledge-gardener`, `knowledge-migrator`, `documenter`, `technical-writer`. ## Per-role findings ### `knowledge-curator` - **Add** a schema-evolution/migration rule per schema change. *(Schema evolution)* https://en.wikipedia.org/wiki/Schema_evolution - **Add** controlled-vocabulary/ontology governance. *(Ontology, information science)* https://en.wikipedia.org/wiki/Ontology_(information_science) - **Add** linking methodology (atomic, linked pages). *(Zettelkasten)* https://zettelkasten.de/overview/ ### `knowledge-ingestor` - **Add** an extraction/summarization methodology (entities/relations; progressive summarization). *(Information extraction)* https://en.wikipedia.org/wiki/Information_extraction ; *(Automatic summarization)* https://en.wikipedia.org/wiki/Automatic_summarization - **Add** an atomic-note method (one idea per page). *(Zettelkasten, Create a Zettel from Reading Notes)* https://zettelkasten.de/posts/create-zettel-from-reading-notes/ - **Add** source provenance per statement. ### `knowledge-querier` - **Add** a retrieval strategy (avoid "lost in the middle" by retrieving, not dumping long context). *(Lost in the Middle, arXiv 2307.03172)* https://arxiv.org/abs/2307.03172 - **Enforce** grounding/citations. *(RAG, arXiv 2005.11401)* https://arxiv.org/abs/2005.11401 - **Add** an answer-quality eval. ### `knowledge-linter` - **Define** orphan criteria + linking suggestion. *(Wikipedia:Orphan)* https://en.wikipedia.org/wiki/Wikipedia:Orphan - **Add** link-rot handling (archive/replace). *(Link rot)* https://en.wikipedia.org/wiki/Link_rot - **Add** a claim-review date / expiry. ### `knowledge-indexer` - **Add** IA catalog rules (organization/navigation). *(Information architecture)* https://en.wikipedia.org/wiki/Information_architecture - **Keep** index (state) and log (events) strictly separate. *(Digital Garden)* https://maggieappleton.com/garden-history - **Add** a completeness/staleness check. ### `knowledge-gardener` - **Add** a frontmatter/metadata convention. *(Obsidian, YAML front matter)* https://help.obsidian.md/Advanced+topics/YAML+front+matter - **Add** tag taxonomy rules. *(Obsidian, Tags)* https://help.obsidian.md/Editing+and+formatting/Tags - **Add** link repair strategy. *(Link rot)* https://en.wikipedia.org/wiki/Link_rot ### `knowledge-migrator` - **Add** OKF/Frictionless conformance. *(Open Knowledge Foundation)* https://okfn.org/ ; *(Frictionless Data Specifications)* https://specs.frictionlessdata.io/ - **Add** migration idempotency/no-data-loss check. - **Structure** content before migrating. ### `documenter` - **Add** Diátaxis content typing. *(Diátaxis)* https://diataxis.fr/ - **Structure** architecture.md with C4 views. *(C4 model)* https://c4model.com/ - **Apply** arc42 + README conventions. *(arc42)* https://arc42.org/overview/ ### `technical-writer` - **Add** a style guide. *(Google developer documentation style guide)* https://developers.google.com/style ; *(Microsoft Writing Style Guide)* https://learn.microsoft.com/en-us/style-guide/welcome/ - **Separate** tutorial/how-to/reference per Diátaxis. *(Diátaxis, Start here)* https://diataxis.fr/start-here/ - **Generate** API reference from the OpenAPI spec. *(OpenAPI)* https://spec.openapis.org/oas/latest.html