The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data from project manifest files and lockfiles, which creates a potential surface for indirect prompt injection attacks where maliciously crafted metadata in those files could attempt to influence the agent's behavior.
Ingestion points: Dependency manifests (e.g., package.json, requirements.txt, pyproject.toml, Cargo.toml, go.mod) and lockfiles are parsed as primary input to the audit workflow.
Boundary markers: The instructions do not define explicit delimiters or warnings to ignore embedded instructions within the processed project files.
Capability inventory: The skill utilizes file system access to read project configurations and subprocess execution to run auditing tools such as npm, pip, cargo, snyk, and trivy.
Sanitization: The workflow relies on the specialized auditing tools to handle the data; no explicit sanitization or validation of the dependency file content is performed by the skill instructions themselves.

dependency-audit