The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted clinical data for report generation, which presents an attack surface for indirect prompt injection.
Ingestion points: Multiple scripts (e.g., scripts/extract_clinical_data.py, scripts/validate_case_report.py, and scripts/check_deidentification.py) are designed to read and parse the content of clinical reports provided by the user.
Boundary markers: The skill instructions strongly emphasize HIPAA de-identification and CARE guideline compliance, but they do not provide explicit delimiters or instructions for the agent to ignore embedded commands within the processed clinical text.
Capability inventory: The skill possesses file system read/write capabilities via its provided Python scripts. It does not exhibit network access in the provided code.
Sanitization: The skill utilizes regex-based extraction and validation (e.g., terminology_validator.py and check_deidentification.py) to sanitize specific medical entities and PII, which serves as a partial mitigation against malformed data.
[COMMAND_EXECUTION]: The SKILL.md file provides instructions for executing a shell command to generate schematics (python scripts/generate_schematic.py).
Evidence: The command python scripts/generate_schematic.py "your diagram description" -o figures/output.png is documented as a mandatory step for report generation. Although the script generate_schematic.py is missing from the provided file list, the instructions encourage the agent to execute shell commands based on natural language descriptions, which could be exploited if the underlying script or the agent's shell integration lacks sufficient input sanitization.

clinical-reports