The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git rev-parse --show-toplevel and git remote -v to detect the project stack and environment. These commands are localized to the reconnaissance phase and are necessary for the tool's auditing functionality.
[DATA_EXPOSURE]: The instructions direct the agent to scan sensitive files like .env, secrets, and configuration manifests. This is the primary intended use case for a security audit tool and is performed to identify vulnerabilities rather than to exfiltrate data.
[PROMPT_INJECTION]: The skill processes untrusted external data (the user's source code and configuration files), which constitutes an indirect prompt injection surface.
Ingestion points: Reads files such as package.json, pyproject.toml, .env, and Dockerfile from the target project.
Boundary markers: None explicitly defined for data ingestion.
Capability inventory: Includes file-read access and the execution of basic Git CLI commands.
Sanitization: The skill includes a 'VERIFY' step requiring the agent to cross-reference findings with library documentation and project constraints before reporting, which acts as a logical control against false positives.

ralph-security