The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external manifest files (such as package.json, requirements.txt, Cargo.toml, etc.), which represents a potential indirect prompt injection surface where malicious instructions could be embedded in metadata.
Ingestion points: The skill explicitly instructs the agent to scan manifests including package.json, requirements.txt, Cargo.toml, and go.mod as noted in the 'Core Workflow' section of SKILL.md.
Boundary markers: The instructions do not provide specific delimiters or 'ignore' directives to help the agent distinguish between manifest data and its own system instructions.
Capability inventory: The skill provides a list of commands for package managers (npm, yarn, pnpm, pip) which allow for updating, installing, and auditing packages.
Sanitization: There are no specific instructions for sanitizing, validating, or escaping the content of the manifest files before they are processed by the agent.

dependency-doctor