The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes user-provided clinical trial data and incorporates it into generated results. Malicious instructions within these inputs could theoretically influence the agent's downstream tasks.
Ingestion points: User inputs for indication, phase, population, and criteria provided in external files are ingested by scripts/main.py.
Boundary markers: Present; the skill instructions in SKILL.md require the agent to validate the objective and scope before execution.
Capability inventory: scripts/main.py includes capabilities for file writing (json.dump) and CLI-based processing.
Sanitization: Absent; there is no specific logic in scripts/main.py to sanitize or filter the input strings for natural language instructions before they are interpolated into criteria templates.

inclusion-criteria-gen