Skill chains Team plan
Link 2 to 12 skills into an ordered chain and draw it in an editor. Your agent runs it as one skill, and each step is verified like any other read.
Anyone can copy a skill file from a repo into their agent. SkilVault adds what that leaves out: a signature your machine verifies, a review before anything ships, and tools that keep a growing library usable.
A skill is instructions your agent follows with your permissions. Where it came from matters more than how it is worded.
SkilVault signs every version offline. The signing key never sits on the server, so a hacked server cannot make your agent read something new.
$ skilvault read changelog-entry-on-completion
skilvault: refusing 'changelog-entry-on-completion' v1.0.1: it is not signed by a trusted keySkills are reviewed for quality and safety before publication.
A new skill or an update is prepared for review.
Quality and safety checks before release.
Your client verifies the approved version before reading.
A published version is never edited. A fix ships as a new, higher version, and an undo is an old version republished as a new one.
Your agent gets a short entry for each skill you use. The full text is fetched, and verified, only when the agent actually needs it.
Free automatically loads your five chosen skills. Paid plans automatically load relevant skills across the library. Every plan can browse and explicitly read any published skill; Free can read 10 new skills a day beyond its five active ones.
Link 2 to 12 skills into an ordered chain and draw it in an editor. Your agent runs it as one skill, and each step is verified like any other read.
Your notes stay in local files, apart from the signed pack, and secret-looking notes are refused. It is guidance for the agent, not enforcement.
Off by default. Turn it on and only scrubbed rules are sent. When enough different accounts report the same thing, a draft update is prepared, and it still needs SkilVault's signature.
A search that finds nothing can become a request. It goes through the same checks and arrives as a draft, never straight into the catalog.
Every account uses an authenticator app or a passkey, with recovery codes. Sensitive actions ask again. API keys are stored only as a hash.
Each plan states its keys, its skills and its hourly downloads. A limit gives you a 429 with a Retry-After, not a surprise.
Choose your skills, connect your machine once, and see the signature check for yourself.