llm-wiki

Use to perform a deep semantic audit of an OKF bundle. Reads all concept docs and identifies contradictions, stale claims, orphan pages, missing cross-references, and concepts mentioned but not written.

Use to find authoritative seed URLs or source datasets for a knowledge gap in an OKF bundle. Given a topic or concept ID that needs enrichment, returns recommended ingest sources.

Use to crawl seed URLs and enrich or mint OKF concept docs from web pages. Manages its own crawl state file and budget. Implements the four-gate reference test and strict augmentation rules.

Use to enrich a single OKF concept doc from raw source metadata. Reads the existing doc (if any), reads raw metadata, composes a conformant OKF document, and writes it via okf_doc.py. One agent per concept; dispatch N in parallel for N concepts.

Use when writing or rewriting a single OKF concept document — a markdown file with YAML frontmatter and a structured body. Covers frontmatter conventions, body structure, cross-linking, citation format, and what NOT to invent.

Use when the user invokes /llm-wiki:enrich or asks to re-enrich specific concept docs. Fans out okf-concept-enricher subagents for the named concepts (or all concepts if none specified).

Use when the user invokes /llm-wiki:index or asks to regenerate the index.md files in an OKF bundle. Runs okf_index.py for deterministic bottom-up index generation.

Use when the user invokes /llm-wiki:ingest or asks to ingest a source into an OKF bundle. Detects source type, routes to the right adapter, runs the supervised ingest loop.

Use when ingesting a BigQuery dataset into an OKF bundle. Covers listing concepts (with shard-family detection), pulling schema/partitioning metadata via okf_bq.py, and routing to the authoring-concepts skill for doc writing.

1 hook across 1 event