The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied technical requirements to generate research directions and reports, creating a surface for injection. * Ingestion points: The requirement_text field in SKILL.md serves as the entry point for external data. * Boundary markers: The workflow in assets/workflow.md and SKILL.md requires distinguishing between source facts and analyst interpretations. The structured JSON payload defined in assets/payload-schema.md enforces data boundaries. * Capability inventory: The skill utilizes network-enabled MCP tools for patent searching and executes a local Python script (scripts/render_report.py) to write report files to the filesystem. * Sanitization: The rendering script performs mandatory HTML escaping on all payload content and validates URLs to exclude embedded credentials. It also scans rendered HTML for forbidden fragments like <script> tags or event handlers.
[COMMAND_EXECUTION]: The skill executes a bundled Python script, scripts/render_report.py, for data validation and report generation. * The script is a static component of the skill package and operates on local file paths defined in assets/paths.md. * It incorporates defensive checks against overwriting files without authorization and refuses to process symbolic links.

identify-rd-directions-rd