healthcare-mcp-hardened

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements explicit guardrails to protect patient confidentiality, forbidding the transmission of clinical data to external network commands and limiting file output to current working directories or user-specified private folders.
  • [COMMAND_EXECUTION]: The skill uses shell commands to execute a local FHIR client script as described in SKILL.md. This usage is mitigated by safety instructions that prevent the chaining of commands for data exfiltration.
  • [PROMPT_INJECTION]: The skill manages clinical data from external FHIR endpoints, which is a potential surface for indirect prompt injection.
  • Ingestion points: FHIR resources are fetched from external servers through the fhir_client.py script (SKILL.md).
  • Boundary markers: The skill provides behavioral safety rules but lacks structural delimiters for the FHIR data payload.
  • Capability inventory: Includes run_shell_command and read_file (SKILL.md).
  • Sanitization: No explicit sanitization or structural validation of retrieved FHIR data is specified in the skill body.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 06:31 PM
Security Audit — agent-trust-hub — healthcare-mcp-hardened