setup-ib

Setup Infinite Brain (ib) skills

Scaffold the per-vault configuration that the ib skills assume:

• Operating block — an ## Infinite Brain vault section in CLAUDE.md (or AGENTS.md) so every agent knows this directory is a typed-node / typed-edge knowledge graph, and which /-commands operate on it.
• _system/ entry points — AGENTS.md (taxonomy, visibility model, frontmatter schema, prohibited actions) and INDEX.md (the master node index every new node must update).
• Defaults — the starting namespace, the default node visibility, and the document language.

This is a prompt-driven skill, not a deterministic script. Explore, present what you found, confirm with the user, then write. Reference: JotaSXBR/obsidian-infinite-brain.

This skill is the entry point for setting up ib in any directory — fresh or existing. /init-vault is the pure scaffolder it delegates to for stamping out the folder structure; don't point users at /init-vault directly for setup.

Process

1. Explore

Look at the current directory to understand its starting state. Read whatever exists; don't assume:

• CLAUDE.md and AGENTS.md at the root — does either exist? Is there already an ## Infinite Brain vault (or ## Agent skills) section?
• _system/ — does it exist? Which files are present (AGENTS.md, INDEX.md, NODE-TYPES.md, EDGE-TYPES.md, FRONTMATTER-SCHEMA.md)?
• Node-type folders at the root (pillars/, decisions/, concepts/, … raw/) — is this already a scaffolded vault, a partial one, or a plain repo?
• _templates/Template - Infinite Node.md — has a vault template been created?
• git remote -v — is this a code repo that also wants a vault, vs. a dedicated vault directory?

Classify the directory as one of: (a) already a complete vault, (b) partial (some _system/ or folders), or (c) not a vault yet.

2. Present findings and ask

Summarise what's present and what's missing. Then walk the user through the decisions one at a time — present a section, get the answer, move to the next. Don't dump them all at once.

Assume the user may not know these terms. Each section opens with a short explainer (what it is, why the ib skills need it, what changes with a different choice), then the choices and the default.

Section A — Vault location & scaffolding.

Explainer: The ib skills read and write typed nodes under folders like pillars/, decisions/, concepts/, and a _system/ directory that holds the taxonomy and the master index. They need a place that is the vault. In a dedicated vault directory that's the root; in an existing code repo you usually don't want 17 node folders at the root, so the vault lives in a subfolder.

• Root — this directory is the vault. Best for a dedicated vault repo. (Default if no src/, package.json, etc. is present.)
• Subfolder (e.g. vault/ or brain/) — the vault lives in a subfolder of an existing project. Ask for the folder name.

Then, based on the classification from step 1:

• If not a vault yet or partial, offer to run /init-vault in the chosen location to scaffold the full structure (17 folders, _system/ files, template, two example nodes). If the user declines, seed only the minimum: _system/AGENTS.md and _system/INDEX.md.
• If already complete, skip scaffolding — only the CLAUDE.md block and defaults need writing.

Section B — Namespace.

Explainer: Every node carries a namespace (e.g. personal, work, research) that groups it under a top-level area. The vault can hold several namespaces, but the skills need a default to stamp on new nodes when you don't specify one.

Default: personal. Ask the user for their starting namespace.

Section C — Default visibility.

Explainer: Every node carries a visibility — public (safe across namespaces and general answers), namespace (used only when the task matches the node's namespace), private (used only when the human explicitly asks for that private context), or system (agent-facing operating rules, never presented as user content). The skills apply this default to new nodes.

Default: public (use namespace if the vault will mix several areas — upstream recommends namespace when unsure). Confirm or override.

Section D — Document language.

Explainer: This sets the language the skills write human-readable node content in — titles, summaries, body prose, edge note fields, and the text of query answers and health reports. It does not change the vault's structural vocabulary: frontmatter field keys (id, title, type, …), the 17 node types, the 10 edge types, the visibility values (public/namespace/private/system), id slugs, namespace names, tags, folder names, and skill/command names always stay in their canonical English form so cross-references, file paths, and the taxonomy remain stable. Choosing a non-English language means everything except that fixed terminology is written in the chosen language.

Default: English. Ask the user which language node content should be written in (e.g. English, 한국어 (Korean)). Record the answer as the document language.

3. Confirm and edit

Show the user a draft of:

• The ## Infinite Brain vault block to add to whichever of CLAUDE.md / AGENTS.md is being edited (see step 4), with <namespace>, <visibility>, and <language> filled in.
• The list of _system/ files that will be created (if any), or "already present — leaving as-is".

Let them edit before writing.

4. Write

Pick the file to edit:

• If CLAUDE.md exists, edit it.
• Else if AGENTS.md exists, edit it.
• If neither exists, ask the user which one to create — don't pick for them.

Never create AGENTS.md when CLAUDE.md already exists (or vice versa) — always edit the one that's already there. If an ## Infinite Brain vault block already exists in the chosen file, update its contents in place rather than appending a duplicate. Don't overwrite the user's edits to surrounding sections.

Write the block using the seed in this skill folder as a starting point — vault-claude-block.md — substituting the namespace, visibility, and document language from Section B/C/D. When the vault lives in a subfolder, note the path in the block's first line ("The vault/ directory is an AI-first knowledge-graph vault…").

Then ensure the _system/ entry points exist:

• If the user opted into /init-vault, invoke it in the chosen location, passing along the namespace from Section B and the document language from Section D so it doesn't re-ask, and skip its operating-block step — the CLAUDE.md/AGENTS.md block is this skill's job and was just written above.
• Otherwise seed the minimum so the other ib skills have something to read: _system/AGENTS.md and _system/INDEX.md (empty per-type tables). For AGENTS.md, copy the seed in this skill folder — vault-agents-template.md — substituting <namespace> from Section B and <language> from Section D; it carries the full operating rules (taxonomy, visibility model, frontmatter schema, document-language rule, log-writing rules, prohibited actions, first-session protocol). Leave any existing _system/ file untouched.

5. Done

Tell the user setup is complete and which ib skills now have the context they need (init-vault, convert-note, query-vault, organize-vault, vault-health). Mention they can edit the ## Infinite Brain vault block or _system/*.md directly later — re-running this skill is only needed to change the namespace/visibility/language defaults or relocate the vault.