obsidian-llm-wiki

The heavy, infrequent (~monthly) maintenance pass for the wiki — a superset of `lint`. Use on /deep-lint, "monthly maintenance", "deep clean the wiki", "audit confidence", or "check my sources are up to date". Does everything `lint` does (dead links, orphans, unindexed pages, conflicts) PLUS reconciling query-time `flagged:` freshness flags, confidence coverage & correctness (changed + flagged + a sampled cold tail — never a full-vault LLM re-read), staleness scoring, capped freshness probes against the original ONLINE sources, the IDEAS.md Monitor review (its sole standing delegation), and a qmd refresh if enabled. Token-bounded by design; run ~monthly or when flags accumulate. Applies fixes only after confirming large or uncertain changes.

Sync THIS LLM-Wiki framework with its public GitHub repo, ONE direction per run: --push (vault → repo) publishes your framework; --pull (repo → vault) updates your framework from a newer repo version. Also builds a standalone content-free copy. Use when the user says "publish/export the framework", "update the public repo", "pull the latest framework", or "make the base template". Push shows the diff and commits+pushes only after you confirm; pull previews first and writes nothing until --apply. Ships the engine (CLAUDE.md, Manual.md, skills, graph config) + a tracked demo in examples/seed/, an MIT LICENSE + README + setup.sh + a .gitignore that TRACKS the framework and IGNORES all knowledge; strips all wiki/raw content + personal data; keeps the AI/LLM-research specialisation. Never copies your knowledge; the build dir is deleted after publishing. Full runbook: the RUNBOOK.md in this skill folder.

Health-check the Obsidian wiki — the "static analysis" pass for a knowledge base. Use when the user runs /lint, /health, /scan, or asks to "check the wiki", "find broken links", "clean up the wiki", or "find gaps/orphans/conflicts". Read-only scan that reports dead links, dead media embeds, orphan pages, pages missing from index.md, unresolved knowledge conflicts, and the count of pending `flagged:` freshness flags (≥5 suggests a deep-lint). Also restores the graph colour palette on request (on-demand only, never on a routine scan). Proposes fixes but only applies them after the user confirms.

Produce a user-facing DELIVERABLE (report, brief, literature review, slide deck, table, email, outline, …) into the output/ directory — grounded in the wiki and strictly following the user's instruction. Use when the user runs /output, or says "write/draft/produce/generate me a <X> about <Y>", "make a <report/brief/deck/email/table> from my notes/wiki", or "turn [[topic]] into <format> in output/". Distinct from `query` (answers inline or files a cited synthesis INTO the wiki) and `export-okf` (exports the whole vault). The deliverable goes to output/, cites the wiki pages it draws on, clearly labels any general-knowledge content, and NEVER fabricates facts, figures, quotes or citations.

Answer questions against the local Obsidian wiki — not from model memory. Use when the user runs /query, or asks in natural language about "my notes / my wiki / what I've researched / my past decisions / what do I know about X". Always read wiki/index.md FIRST to locate pages, then read them in full, then answer with [[wikilink]] citations. If the wiki has nothing relevant, say so explicitly before giving any general-knowledge answer. Offers to file high-value answers back into wiki/syntheses/ so explorations compound.

OPTIONAL, DORMANT semantic-search layer over the wiki, powered by qmd (local hybrid BM25 + vector + rerank). Use only when the wiki has outgrown index.md AND the user has installed + enabled qmd. The agent auto-detects qmd; if it is absent or disabled this skill does nothing and the normal index.md → grep search runs unchanged. Provides confidence-aware semantic retrieval as the query/output fallback, plus a refresh hook that updates a page's qmd embedding whenever it is created or changed. Retrieval only — never writes or decides. Trigger: /qmd-search <query>, or internally as the query/output semantic fallback.

Compile inbox files in raw/ into the wiki (source + entity/concept/model/benchmark pages), update index + log, then sort the raw file into its category subfolder. Use on /ingest, when files are dropped into raw/, or "add this to my wiki / ingest this / file this source / process my inbox". Non-.md sources (PDF, PPTX, DOCX, XLSX, image, audio, HTML, CSV, EPUB, YouTube/web URL) are converted to Markdown first; runs a cheap de-dup check; supports standard/concise/research modes (research is opt-in or ask-first). Scans the raw/ ROOT only (or one given path); never edits raw file contents — only relocates them.

Export the wiki/ knowledge base as a conformant Open Knowledge Format (OKF v0.1) bundle — a portable, vendor-neutral directory of Markdown concept files. Use when the user runs /export-okf or asks to "export to OKF", "make an OKF bundle", "produce a portable/shareable bundle", or share the vault with other agents/tools. Deterministic and READ-ONLY w.r.t. the vault (never mutates wiki/); writes the bundle to okf-export/. It derives OKF frontmatter (description/resource/timestamp), converts [[wikilinks]] to bundle-relative markdown links (code-aware), copies referenced media, generates per-directory index.md (§6) + a root log.md (§7), and validates conformance (§9).

Deep Raw-layer capture: collect a whole topic's web sources at once. Given a seed URL (or a few), fetch it, find the relevant links it cites (papers/repos/docs), and — after showing you a preview you approve — capture those too into raw/, with images localised, ready for /ingest. Use when the user runs /gather, or says "gather sources on X", "deep-capture these links", "grab this article AND the repos/ papers it cites". Safe by default (conservative caps + preview + a hard ceiling) for non-experts; fully overridable (depth/pages/include/exclude/--yes) for experts. Opt-in; does nothing until invoked. It only CAPTURES into raw/ — the normal /ingest skill then compiles. Read-only except for adding new raw files (never edits existing raw content).