Install AutoVault.
One command. No daemon.

Step 1 · Install

All three channels install the same release, so pick whichever you already use. The shell installer does the most: it creates ~/.autovault, leaves the folder owned by you, and seeds the bundled skills — set AUTOVAULT_NO_BOOTSTRAP=1 if you would rather start empty. Nothing runs in the background; your agent starts AutoVault only when it needs it.

# Node 24+: install the published npm package
$ npm install -g @autoworks-ai/autovault
# Or run the installer script end-to-end
$ curl -fsSL https://autovault.sh | sh
# macOS: also available through the tap
$ brew install autoworks-ai/tap/autovault
$ autovault skill list

Rather have your agent do it?

If you want Claude Code to configure its own AutoVault bootstrap skill, give it this prompt. The hosted skill is a raw SKILL.md; the agent should fetch it, show you the behavior, install it locally only after approval, then run it.

Fetch https://autovault.dev/skill.md, show me what it will do, install it into ~/.claude/skills/autovault-bootstrap/SKILL.md if approved, then run /autovault-bootstrap.

Finish setup

autovault setup finds the agents already on your machine — it looks in ~/.claude/skills, ~/.codex/skills and ~/.cursor/skills — then walks you through each skill it found and asks what you want done with it. Safe to run again whenever you install a new agent.

$ autovault setup --review

Check it worked

One command tells you whether the CLI, the vault folder, your signing key, and agent discovery are all in place.

Step 2 · Add your first skill

Point autovault add at a folder, a GitHub repo, an agentskills slug, or a direct SKILL.md URL. Wherever it came from, it goes through the same gate before it reaches your vault — the frontmatter is repaired, the contents are scanned, what the skill declares is checked against what it actually does, and the result is signed.

$ autovault add ./skills/skill-author --sync-profiles --yes

The vault is a folder

The final model is intentionally boring: a regular folder on disk. This is the current implementation layout: SQLite index, local signing key, source skills, source metadata, signed manifests, rendered variants, and profile links. Select a row to inspect how the folder is read.

How skills reach each agent

Every agent reads skills from its own folder. Syncing links the skills you have admitted into those folders, and never overwrites a file you put there yourself. Which skills go where comes from each skill's own metadata, or from named profiles in profiles.config.json. Remote clients read over MCP rather than expecting symlinks on the machine.

$ autovault sync-profiles --discover

Step 3 · Run it from your agent

The same skill is now validated, scoped, and rendered for each target agent. The skill name stays stable, while tool names are transformed to match the caller.

> use skill-author to draft a new SKILL.md
✓ tool resolved: Read, Edit, Write
# extracting...
 "This 24-page report covers Q1 platform metrics, with three..."

Remove a skill cleanly

If a skill no longer belongs in the vault, remove it through AutoVault rather than deleting folders by hand. The command deletes the vaulted copy and its vault-local transforms, refreshes generated profiles, and prunes AutoVault-managed symlinks from discovered native agent roots.

$ autovault remove skill-author
# automation-friendly output
$ autovault remove skill-author --json

Where to next