The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_schematic.py utilizes the subprocess.run function to execute an internal script (scripts/generate_schematic_ai.py). While the command is constructed using a list of arguments to mitigate shell injection, it still represents a capability for local process execution triggered by user input.
[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection through the processing of user-supplied medical documents.
Ingestion points: Data is read from external .tex and .txt files in scripts/check_completeness.py, scripts/validate_treatment_plan.py, and scripts/timeline_generator.py.
Boundary markers: Absent. No delimiters or instructions are used to separate clinical data from potential instructions.
Capability inventory: The skill can execute local commands in scripts/generate_schematic.py, write to the file system in scripts/generate_template.py and scripts/timeline_generator.py, and perform network requests to an external AI service in scripts/generate_schematic_ai.py.
Sanitization: Absent. The ingested data is processed directly using regex patterns without escaping or validation against malicious instruction patterns.
[EXTERNAL_DOWNLOADS]: The scripts/generate_schematic_ai.py script communicates with the OpenRouter API (https://openrouter.ai/api/v1) to generate and review scientific schematics. This involves sending prompts and image data to a third-party service, which is a documented and expected behavior for the skill's AI features.

treatment-plans