skill-scanner

To scan agent skills for prompt injection and malicious code, you can use a static analysis scanner to audit SKILL.md files and bundled scripts. This process detects hardcoded secrets, dangerous shell commands, and excessive tool permissions before execution.

Yes, you need the uv CLI for Python package management to run the static analysis scanner. It requires the pyyaml dependency to parse configuration files and evaluate agent skill directories for security vulnerabilities.

Supply chain protection for agent skills is the process of identifying suspicious remote URL fetching and hidden structural attacks like symlink exfiltration or npm lifecycle hooks. It prevents untrusted skills from compromising your system during execution.

To audit tool permissions for the principle of least privilege, you run a static analysis scanner on the target skill directory. The scanner evaluates requested tool access against minimal requirements and provides structured JSON output detailing over-permissioned agents.

Yes, you can get structured JSON output from an agent skill security audit. The scanner evaluates SKILL.md files and bundled scripts, returning structured JSON data that details detected prompt injection patterns, credential exposure, and excessive permission requests.

A skill scanner can detect security vulnerabilities including prompt injection patterns, hardcoded secrets, dangerous shell commands, and excessive tool permissions. It also identifies suspicious remote URL fetching and structural attacks like symlink exfiltration.