skilltotal

SkillTotal is a security analysis tool for AI components. It scans an AI component before it is trusted and reports supply-chain risk, dangerous capabilities, prompt-injection, and exfiltration surfaces, all derived from the component itself.

Its analysis is deterministic and static, with no execution and no LLM involvement. A submission can be a public git URL from GitHub, GitLab, Bitbucket, or Hugging Face, or an npm:/pypi: package. The supported component types named on the site include MCP servers, agent skills or plugins, model repositories, and packages. The resulting report includes a risk score, all capabilities the component exposes, a behavioral trait fingerprint mapped to the Cloud Security Alliance agentic threat model, MAESTRO, and MITRE ATLAS, and findings anchored to file, line, and snippet evidence. Reports can be exported as JSON or SARIF.

The service supports scanning multiple servers by pasting a configuration, and it can scan a skill or project from an uploaded ZIP file. It also states that nothing is stored, public reports are cached and shared, and scans on the site run on the provider’s servers without executing the submitted code.

SkillTotal is free and requires no account. It is open source under the Apache-2.0 license, and it can also be run locally so that code never leaves the machine. The page names GitHub and PyPI, and mentions a command-line install path through pipx install skilltotal.