The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to systematically read and analyze the project's source code in the src/ directory. This creates an attack surface where a malicious file within the audited codebase could contain instructions intended to manipulate the agent's audit results or behavior.
Ingestion points: Source code files in the src/ directory and project configuration files.
Boundary markers: The instructions do not define delimiters or specific warnings to ignore instructions found within the audited code.
Capability inventory: The agent uses tools to read files and execute shell commands (grep, npm audit).
Sanitization: No sanitization or filtering of the source code content is specified before the agent processes it.
[REMOTE_CODE_EXECUTION]: The skill suggests running npx is-website-vulnerable, which downloads and executes a package from the npm registry. While this is a common method for running security audit tools, it involves the execution of code retrieved from a remote registry at runtime.
[COMMAND_EXECUTION]: The skill employs shell commands to perform the audit, such as grep for identifying leaked secrets (e.g., Stripe or Supabase keys) and npm audit or pnpm audit to check for known vulnerabilities in the project's dependencies.

security-audit