TankSecurity-first package manager for AI agent skills
Today's skill registries have no versioning, no lockfiles, no permissions, and no code signing. Tank is the npm for agent skills — with security built in from day one.
Declare what your agent is allowed to do
skills.json is like package.json — but with a permissions field. If any skill exceeds the budget, installation fails.
{
"skills": {
"@vercel/next-skill": "^2.1.0",
"@community/seo-audit": "3.0.0"
},
"permissions": {
"network": {
"outbound": ["*.anthropic.com"]
},
"filesystem": {
"read": ["./src/**"],
"write": ["./output/**"]
},
"subprocess": false
}
}Security at every layer
From publish to install to runtime — every step is verified, signed, and sandboxed.
Every skill is pinned with sha512 hashes in skills.lock. If content doesn't match its hash, installation fails. No silent tampering.
Declare what your agent is allowed to do — network, filesystem, subprocess. If any skill exceeds the budget, installation fails.
Transparent 0–10 security score for every skill. Code signing, static analysis, reproducible builds, vulnerability checks — all visible.
Not a social contract — enforced. A patch that adds network access? Rejected. Breaking changes require a major bump.
Install, publish, audit, and manage permissions from the terminal. Designed for developers, not dashboards.
Built in the open from day one. MIT licensed. Community-driven security reviews. No vendor lock-in.
Get started in seconds
Everything you need, right from the terminal.
# Install with full integrity verification
$ tank install @vercel/next-skill
# See exactly what your agent is allowed to do
$ tank permissions
# Check signatures, vulnerabilities, permission escalations
$ tank audit
# Publish with mandatory signing and static analysis
$ tank publish
Agent skills are more dangerous than npm packages
A malicious npm package runs inside your app's sandbox. A malicious agent skill runs with the agent's full permissions — reading any file, making API calls with your credentials, executing shell commands. The attack surface is fundamentally larger.
Ready to secure your agent skills?
Tank is open source and free. Start publishing and installing skills with confidence.