The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is explicitly designed to locate and inventory credentials stored in sensitive system paths, including ~/.aws/credentials, ~/.kube/config, ~/.ssh/, and browser profile databases. This is the primary and legitimate function of the tool for security auditing purposes.
[COMMAND_EXECUTION]: The skill utilizes the ggshield CLI to perform system-wide audits and manages functional plugins such as machine_scan via the ggshield plugin install and enable commands.
[DATA_EXFILTRATION]: The skill provides functionality to generate and upload a machine-wide inventory of credentials to the GitGuardian dashboard using the ggshield machine inventory command. These operations target official vendor domains (gitguardian.com) and are intended for organizational governance.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by scanning the contents of files across the entire home directory, creating a vulnerability surface for indirect instructions.
Ingestion points: The ggshield machine scan command reads all files within the $HOME directory, including repository contents and configuration files (SKILL.md).
Boundary markers: The instructions direct the agent to report findings by metadata (file, line, type, validity) and explicitly forbid logging raw secret values, providing a layer of oversight.
Capability inventory: The skill has the capability to execute shell commands via ggshield and perform network operations to communicate with the GitGuardian API (SKILL.md).
Sanitization: Content sanitization is not explicitly defined in the markdown instructions and is handled by the underlying ggshield security engine.

scan-machine