The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/generate_schematic.py to execute its internal AI generation script. This is handled using a list of arguments rather than a shell string, which prevents command injection.
[EXTERNAL_DOWNLOADS]: The skill communicates with the OpenRouter API (openrouter.ai) to generate and perform quality reviews of scientific schematics. This involves sending prompt and image data to a well-known external service. This is a documented functional requirement and uses standard API practices.
[INDIRECT_PROMPT_INJECTION]: The skill processes clinical data provided by users, which creates a surface for potential indirect prompt injection.
Ingestion points: Clinical data is read and processed in scripts/validate_case_report.py, scripts/check_deidentification.py, and scripts/extract_clinical_data.py.
Boundary markers: The skill does not use explicit boundary markers when interpolating clinical text into prompts.
Capability inventory: The skill has access to Read, Write, Edit, and Bash tools.
Sanitization: No active sanitization is performed on input text, although scripts/check_deidentification.py provides regex-based scanning to help users identify and remove sensitive data.

clinical-reports