The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly instructs the agent to execute a local script using the Bash tool: python scripts/generate_schematic.py "your diagram description" -o figures/output.png. Because the command interpolates a natural language description provided by the user into a shell execution, it is vulnerable to command injection if the script does not properly sanitize its input arguments.
[DYNAMIC_EXECUTION]: The skill relies on a suite of local Python scripts located in the scripts/ directory to perform core functions including generate_schematic.py, validate_case_report.py, validate_trial_report.py, check_deidentification.py, format_adverse_events.py, generate_report_template.py, extract_clinical_data.py, compliance_checker.py, and terminology_validator.py. The execution of these scripts at runtime represents a dynamic code execution surface.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted medical data such as patient histories, laboratory results, and clinical trial records.
Ingestion points: Clinical data processed in SKILL.md (e.g., SOAP notes, H&P, and diagnostic reports).
Boundary markers: The instructions do not specify any delimiters or safety markers to differentiate between medical data and agent instructions.
Capability inventory: The skill is granted powerful tools including Bash, Write, Edit, and Read, which could be abused if malicious instructions are embedded in the clinical data.
Sanitization: While the skill highlights HIPAA de-identification (removing personal identifiers), it lacks guidance for sanitizing data to prevent the execution of embedded prompts or instructions.

clinical-reports