The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes docker run commands to initiate security scans. This is the core functionality of the skill and is implemented with safety controls, including the --rm flag for temporary containers and the :ro (read-only) flag for volume mounts to prevent modification of the host filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of third-party skill code and metadata during the scanning process.
Ingestion points: Input is provided via local directory paths or remote Git URLs passed as arguments to the scanner.
Boundary markers: The skill mandates the use of Docker for process isolation and read-only mounts to ensure the scanned content cannot affect the execution environment.
Capability inventory: The skill's capabilities are limited to executing the docker CLI for analysis.
Sanitization: Security is maintained by performing static analysis (parsing code structures and checking signatures) rather than executing the analyzed code, which prevents the untrusted input from gaining active control.

skill-scan