open-knowledge

OpenKnowledge — agent guidance

OpenKnowledge (OK) is a markdown-CRDT collaboration platform exposed via MCP. This skill is the single source of OK agent guidance. Every rule below is a MUST unless marked otherwise. Depth lives in references/*.md — one level deep; load a reference when its task comes up.

Skill version tracks @inkeep/open-knowledge-server. cat ~/.ok/skill-state.yml shows what's installed. ok seed needs @inkeep/open-knowledge >= 0.4.0; if it errors unknown command, npm install -g @inkeep/open-knowledge.

Setup (not connected yet?). If the mcp__open-knowledge__* tools aren't available in your client, this project isn't wired up on this machine — see references/setup.md for the rung ladder (approve .mcp.json → ok start CLI → optional desktop app) and the canonical quickstart.

TL;DR — the 90% case

1. Reads: exec("cat …") for one doc, exec("ls -A …") for a directory (folder defaults + template menu), exec("grep …") for literal, search for ranked retrieval. Native Read / Grep only on source code (.ts / .py / …), never on in-scope .md / .mdx.
2. Writes: write({ document: { path, content } }) for a new or full-replace doc; edit({ document: { path, find, replace } }) for a body find/replace; edit({ document: { path, frontmatter } }) for a frontmatter merge-patch (null deletes a key). delete({ document }) removes, move({ from, to }) moves/renames. Body find/replace is body-only. Pass a one-line summary (≤80 chars, user-facing outcome) on every content write.
3. Preview / open a doc — determine your ONE surface FIRST (once per session). Stop at first match: OK_DESKTOP_TERMINAL or OK_HOSTED_AGENT set → you're inside OpenKnowledge (desktop terminal / in-app agent panel) → ok open <name> (switches the window the user is already looking at); never paste a localhost URL into your reply here · in-app browser (Claude Code Desktop's Browser pane, Cursor, Codex) → preview_url, then open/navigate it to the doc · else plain CLI → ok open <name>. ok open <name> opens a doc or folder (auto-detected); --skill <name> for a skill. The previewUrl field is a route id, not your open mechanism. Don't preview_screenshot to confirm edits. Full Step-0 procedure + per-surface how-to: references/preview.md.
4. Knowledge layers: capturing a source (ingest), synthesizing findings (research), promoting a decision (consolidate) — procedures, not tool calls; there is no ingest tool. Ingest ships here (references/ingest-and-sources.md); research + consolidate come with the knowledge-base pack. Layer model + packs: references/starter-packs.md.
5. Direct questions: a plain business question ("which customers…", "what did we decide about…") routes to search / exec + a cited chat answer — no "research" keyword needed. Persist only when durable + multi-doc + not already covered, and offer first. See references/corpus-qa.md.
6. Authoring or improving a skill ("write/make a skill", "improve this skill", "turn this into a skill"): STOP and invoke the /open-knowledge-write-skill skill — it owns scope choice (project vs global), the SKILL.md contract, evaluation, and install; don't improvise here. Author through the skill target (write({ skill })), never a raw path document. Skills are REAL folders under the editor dirs' skills/ (.claude · .cursor · .codex · .github · .opencode · .pi · .agents): ONE is the source, the rest are managed copies/symlinks. Read and edit via the tools (skills, edit({ skill })) — they route to the source. Never hand-edit a non-source copy: unedited copies refresh from the source, and editing one forks it into its own skill that stops refreshing.

Tool index — 21 tools (router; the MCP tool descriptions carry each tool's full contract)

• Reads — exec (primary; read-only cat/ls/grep/… plus frontmatter/backlink/history enrichment; one command or one pipe, not a shell), search (ranked BM25 + recency), history (doc versions), links (kind: backlinks|forward|dead|orphans|hubs|suggest, or an array for one call), skills (search + read: query → skills.sh; omit name to LIST managed (Project + Global); name READs one — by name+scope, never path), config (resolved config), palette (authoring forms + html preview starters + theme tokens; palette({ components }) for JSX schemas), preview_url (browser preview URL on demand), share_link (GitHub-substrate share URL; read-only, errors without a GitHub remote), lint (markdown-lint violations: document for one doc, omit for the project; fix: true with document auto-fixes fixable rules in place — attributed, live in the preview; the rest need edit/write), audit (every lint violation + broken internal link in one read-only report, by source file with lines; path scopes; for link VALIDATION use this, not links).
• Writes — four native CRUD verbs, polymorphic over document / folder / template / skill / asset (pass EXACTLY ONE target, nested under its address key): write (create/overwrite; write({ skill: {…} }) authors a skill as a REAL folder at the project's default skill home — live immediately for that folder's agent), edit (body find/replace/frontmatter merge-patch; no asset), delete (remove), move (move/rename, rewrites referrers; a skill also takes scope/toScope for Project↔Global — history resets, re-install). Output mirrors the input key; the preview envelope (previewUrl, warning) stays top-level. Plus install (WHERE a skill lives: add/remove locations additively — editor ids, agents, or custom roots; mode + convert re-form ONLY the locations named; source moves the real folder. The source folder IS the skill — no "uninstall everywhere"; a skill dies only via delete), import (acquire a skill-dir into add's locations; scripts never run), checkpoint (named version), and restore_version (roll back). A folder's frontmatter is open-shape and self-only (does NOT cascade); templates are what new docs start with.
• Conflicts — conflicts (kind: list|content), resolve_conflict (write a resolution + commit; destructive). See references/conflict-resolution.md.

Self-correcting on misuse: constraints JSON Schema can't express ("exactly one target", "find needs a replace", body-XOR-frontmatter) return isError: true with a one-line corrective shape. Read it and retry with that shape; don't guess.

Tools NOT in OK MCP (your host's): preview_start, preview_screenshot, WebFetch, WebSearch, native Read / Grep / Glob / Edit. The STOP rule governs which you may use on in-scope markdown.

STOP — native tools on in-scope .md / .mdx

Route every in-scope markdown read and write through OK's MCP tools — never your host's native file tools. Native Edit / sed / direct Write on in-scope markdown bypasses the CRDT and loses agent attribution in the shadow repo; native reads skip frontmatter, backlinks, shadow-repo activity, and project git history that OK returns for every matched file. When this workspace has OpenKnowledge MCP configured, do not use native file tools on markdown paths inside the content directory. The ban covers every common rationalization:

• Native Read / Grep / Glob on in-scope .md / .mdx — the original case.
• Bash ls / Bash find / Bash cat on dirs containing in-scope markdown — use exec("ls -A …") / exec("find … -name '*.md'") / exec("cat …"). Native returns bare names; exec returns frontmatter, backlink counts, and recent activity. -A shows hidden entries (.ok/, .okignore) without the ./.. noise.
• Glob patterns that target markdown (**/*.md, specs/**, reports/**, docs/**) — use exec with find or exec("ls -A <dir>").
• Dispatching the Explore / general-purpose subagent for markdown-heavy exploration — subagents use native tools internally and bypass OK. Do markdown exploration yourself via exec / search. Subagents remain appropriate for source-code exploration.
• Native Read / Grep on in-scope markdown inside .ok/ — .ok/ is in-scope; treat its .md / .mdx like any other KB file.
• ls / cat / find on skill folders to discover or read a skill — skills are addressed by name+scope, not by path (a skill can live in any editor dir, the .agents/skills/ hub, or a custom root, with copies elsewhere). Use the skills tool.

Not seeing exec is NOT the escape hatch. Wiring, labels, and tool visibility vary by client; some (notably Codex) defer MCP tools behind lazy discovery. Registration is the test, not top-level-symbol visibility — run tool discovery for open-knowledge first. Detail: references/setup.md.

Escape hatch. Native Read / Grep / Glob on .md / .mdx is allowed only when, after running tool discovery (above), no OpenKnowledge MCP server is registered for this project, or immediately after you actually invoked an MCP call and it failed — then begin a user-visible sentence with OpenKnowledge MCP unavailable:. "Not registered" is a conclusion you may only reach after tool discovery turned it up empty — never from the initial tool list alone. Never use the hatch because you skipped your client's MCP path, didn't see exec as a top-level tool, didn't run tool discovery, or rationalized the skill wasn't necessary.

Source code and non-markdown files (.ts, .py, package.json, …): native Read / Grep / Glob always.

Reads — examples

• Read a file: exec("cat <path>.md") — contents + full enrichment.
• List a directory: exec("ls -A <dir>") — per-child frontmatter, recursive markdown counts, most-recently-updated doc per subdir, the folder's own title/description/tags + templates_available. Prefer -A over plain ls.
• Literal search: exec("grep -rn <term> <dir> | head -5") — matches + enrichment on matched files.
• Ranked search: search({ query }) — title boost + body BM25 + recency; use when picking the best doc, not when listing every occurrence.

Writing

Call write / edit as soon as you have content (route through MCP per the STOP rule).

Persist incrementally — the knowledge base IS your checkpoint (MUST). On any multi-step or long-running task — a research sweep, a multi-source synthesis, a batch of docs — write completed work to the KB as you finish each unit: per section, per source, per doc. Never hold finished findings only in your context waiting for one final write at the end. A rate limit, crash, or context compaction mid-task discards everything still unwritten; work already persisted survives, and you resume by reading the doc back. Create the target doc early (skeleton + frontmatter), then edit each section in as it firms up.

Pass a summary on every content write (SHOULD) — a one-line (≤80 char) user-facing note; it becomes the timeline entry. Reach for visual structure (Callout, mermaid, table, html preview) where it carries the point better than prose; call palette as you draft. Advisory write-warnings, MDX authoring, delete/move mechanics, and visual authoring: references/writing.md + references/components-and-visuals.md + references/media-and-assets.md.

Grounding — every factual claim needs a source (MUST)

KB docs are factual artifacts: every claim traceable, and the source lives inside the knowledge base, not on the public web.

Ingest is a procedure, not a tool — binary-vs-text classification, SSRF-safe fetch flags, size + executable gates, wrapper frontmatter — in references/ingest-and-sources.md. Read it before your first capture; a naive fetch-and-paste skips every gate.

• Closed loop. External sources are pulled in by the ingest procedure, then cited locally. A bare [source](https://...) inside a KB doc is not a citation — it is a TODO meaning "still needs ingesting". The chain only works if every leaf is a local doc.
• Every factual claim MUST cite its source at the point of claim. No unsourced speculation.
• Web sources → fetch the page (host WebFetch / WebSearch), ingest it, then cite the path: [source name](./path/to/source.md) (the local doc carries source_url:). Inline [source](URL) is a chat affordance, not a KB one.
• Self-fetched counts. A URL YOU fetched to ground a claim gets the same ingest — no inline-URL downgrade.
• Internal cross-refs → link the OK doc holding the authoritative claim; that doc cites its own sources (chains terminate in preserved local docs).
• No evidence? Search and ingest the result, OR mark (TODO: needs source), OR don't write the claim. Do NOT fabricate — unsourced speculation rots into untraceable tribal lore.

Linking — standard markdown links (MUST)

Link every noun-phrase that names another document — [text](./relative/path.md) — and link liberally. Every link must resolve to a doc that exists by the time you're done (a same-pass forward-reference you create later in the pass is fine; for one that genuinely won't exist, leave the mention as plain prose + a tracked task). Never backtick a link (`[text](./foo.md)` is a bug) and never use HTML <a>. Read brokenLinks on every write/edit response: [] means all links resolve; a populated list names each broken href + reason (no-such-doc / no-such-file / unresolvable) — fix them in a follow-up edit. audit is the authoritative end-state link check (the editor's red-underline is slug-tolerant and lies, so trust the tool). External web sources are NOT inline body links (see Grounding). Full rule set + the [[Page]] legacy note: references/linking.md.

Folders, frontmatter, templates

Every .md / .mdx needs YAML frontmatter — title + description required, tags recommended. Two opt-in, nested folder mechanisms: folder frontmatter (<folder>/.ok/frontmatter.yml — the folder's own open-shape properties; self-only, does NOT cascade into child docs) and templates (<folder>/.ok/templates/ — what new docs start with). Most folders have NO .ok/. A doc's frontmatter is exactly its own on-disk YAML. Structural model + the full pre-write checklist: references/folder-model.md. Template authoring + folder editing: references/template-authoring.md. Frontmatter-vs-body edit rules: references/doc-editing.md.

• Read the folder before writing (MUST). Before creating/editing docs in a folder, call exec("ls -A <folder>") once per folder per session — it returns the folder's title/description/tags + templates_available. Skipping it lands docs that violate folder discipline. (If a folder has no frontmatter AND no templates AND the repo has substantial content elsewhere, it isn't onboarded — run references/onboard-existing-repo.md first.)
• **Use a

...