The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/generate_schematic_ai.py downloads image data and quality reviews from the OpenRouter API (openrouter.ai), which is a well-known service for accessing AI models.
[DATA_EXFILTRATION]: User-provided schematic descriptions are transmitted to the external OpenRouter API. While this is a functional requirement for image generation, users are advised to ensure de-identification of prompts to avoid sending sensitive clinical details to the external service.
[COMMAND_EXECUTION]: The script scripts/generate_schematic.py uses subprocess.run() to execute internal auxiliary Python scripts. The implementation uses list-based arguments which mitigates the risk of shell-injection attacks.
[PROMPT_INJECTION]: Category 8 (Indirect). The skill ingests untrusted clinical data from report files through its validation scripts (check_deidentification.py, validate_case_report.py). Ingestion points: Input file paths provided to CLI scripts. Boundary markers: Absent. Capability inventory: Subprocess execution and network access for schematic generation. Sanitization: The scripts use regex-based extraction but do not explicitly sanitize or escape content before it is processed by the agent.

clinical-reports