clinical-trial-protocol-skill

Fail

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: CRITICALCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Automated security scanners identified a blacklisted malicious URL (manderson.org) within the asset file assets/FDA-Clinical-Protocol-Template.md. The domain is known for phishing and malicious activity, and its presence in a clinical research tool is a significant security risk.
  • [COMMAND_EXECUTION]: In references/04-protocol-operations.md, the skill executes a local Python script using shell commands that interpolate user-provided numerical parameters directly into the argument string without sanitization. This pattern enables arbitrary command injection if an attacker provides crafted strings.
  • [PROMPT_INJECTION]: The skill's architecture for ingesting 'initial_context' documentation lacks security boundaries or sanitization. 1. Ingestion Points: references/00-initialize-intervention.md (initial documentation) and references/02-protocol-foundation.md (custom user templates). 2. Boundary markers: Absent. 3. Capability Inventory: Shell command execution for file concatenation and local script execution. 4. Sanitization: Absent.
Recommendations
  • AI detected serious security threats
  • Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Mar 28, 2026, 08:18 PM