SDTK-BRAIN — Standalone Second-Brain Vault Shipped

SDTK-BRAIN is a standalone, local-first second-brain vault CLI. It is deliberately not one of the six SDTK toolkits, not a pipeline stage, and not installed by sdtk-kit — you install it on its own.

• CLI: sdtk-brain (package sdtk-brain-kit) — standalone, not in the umbrella
• Full command list: sdtk-brain reference

npm install -g sdtk-brain-kit   # separate from `npm install -g sdtk-kit`

SDTK-WIKI or SDTK-BRAIN?

Both build a local knowledge graph, and they share a graph builder and viewer (byte-identity is CI-guarded). Everything else is different — they answer different questions:

	SDTK-WIKI	SDTK-BRAIN
Scope	One product — the repo you are working in	You — personal, across every project
Lives	Inside the repo (.sdtk/wiki); your docs are indexed in place	Its own vault folder, deliberately outside any repo
Job	The memory layer every SDLC phase reads from	A second-brain vault your agent reads and maintains
Content	The repo's own docs, as they already exist	raw/ sources you collect → compiled wiki/ knowledge pages
Install	One of the six, ships in sdtk-kit	Separate: npm i -g sdtk-brain-kit
Q&A	sdtk-wiki ask — cited passages free; synthesized answer Pro	No ask command — your agent does the thinking

Rule of thumb: SDTK-WIKI is the knowledge of one product. SDTK-BRAIN is the knowledge of one person.

The second-brain pipeline verbs (ingest, compile, discover, enrich) used to live in sdtk-wiki; they were removed in wiki 0.10.0 and now belong to SDTK-BRAIN. See the sdtk-wiki reference.

What a vault is

A vault is a standalone folder — one per person, spanning every project, never inside a product repo:

Folder	Owner	Holds
raw/	you	Immutable sources — never rewritten
wiki/	agent + CLI	The compiled markdown knowledge layer
workspace/	you	Scratch space
.brain/	CLI	Machine state, reports, graph

The vault ships its own CLAUDE.md operating contract: your agent is the knowledge engine; the CLI provides only deterministic rails. Everything is local — no LLM in the CLI, no network, no telemetry, no entitlement.

Quick start

mkdir my-vault && cd my-vault
sdtk-brain init                          # scaffold raw/ + wiki/ + workspace/ + CLAUDE.md
cp ~/some-article.md raw/inbox/
sdtk-brain ingest raw/inbox              # semantic extraction (report-first)
sdtk-brain compile --mode safe --apply   # compile into wiki/ knowledge pages
sdtk-brain search "topic"                # deterministic search over the vault
sdtk-brain open                          # docs view + graph view (local viewer)

init refuses a non-empty folder that isn't already a vault — it will never scaffold into a product repo. Re-runs are idempotent and never overwrite your files.

Maintenance is report-first

sdtk-brain lint                          # orphans, broken links, stale claims, contradictions
sdtk-brain maintain --mode safe          # lint + discover + compile preview in one pass
sdtk-brain discover --plan               # gap-analysis plan from wiki evidence
sdtk-brain status                        # vault status

Reports land under .brain/reports; nothing is auto-modified. compile --apply fails closed rather than overwrite a page you or your agent hand-curated — pass --skip-conflicts to land the additive operations and get an explicit list of what it skipped.

Not goals

• Not a pipeline stage — SDTK-BRAIN is not part of idea-to-ship and no SDTK phase depends on it.
• Not in the sdtk-kit umbrella — it installs no SDTK skills and requires no other kit.
• No ask command in the CLI, no cloud, no network, no entitlement.

Where to go next

Command referenceEvery command and option for the vault CLI.

SDTK-WIKIThe project memory layer — the other side of the boundary.

Maturity & versionsExactly what is Shipped, Preview, Pro, or Gated right now.