The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script, scripts/check_no_skill_leak.py, to validate that generated documents do not contain internal skill identifiers or implementation details before they are presented or uploaded to the Jinkō platform.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied scientific questions and contexts of use to generate regulatory documentation. It implements the following security posture for this attack surface: 1. Ingestion points: User input for Question of Interest and Context of Use in SKILL.md; 2. Boundary markers: Explicit instructions to label assumptions and avoid inventing information; 3. Capability inventory: File writing and execution of a local validation script; 4. Sanitization: A mandatory check using check_no_skill_leak.py to remove internal metadata.
[EXTERNAL_DOWNLOADS]: The skill mentions a prerequisite skill (jinko-sdk-setup) which can be installed from the vendor's GitHub repository (novainsilico/jinko-skills). This is a resource provided by the skill's author for necessary SDK functionality and is treated as a safe vendor resource.
[SAFE]: The skill refers to official regulatory guidance from the FDA website and uses standard project management practices. The included Python script performs a benign, developer-oriented safety check.

jinko-task-ichm15-planning