The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an inherent attack surface for indirect prompt injection due to its core functionality of reading and validating project-specific documentation.
Ingestion points: The scripts/validate-docs.sh utility is designed to scan and read the contents of various project files, including README.md, CHANGELOG.md, and any markdown files located within documentation, API, and runbook directories.
Boundary markers: The skill lacks explicit boundary markers or instructions that would prompt the agent to disregard natural language commands found within the user-provided documentation files it processes.
Capability inventory: The skill environment includes access to Read, Write, Edit, and Bash tools, which could be leveraged if an injection attack successfully manipulates the agent's execution flow during the validation process.
Sanitization: While the validation script uses standard shell tools (grep, sed, find) for structural and content linting, it does not implement specific sanitization techniques to distinguish between legitimate documentation content and potential adversarial instructions.

technical-writer