psique-workflow-clinic

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The 'ClaudeOrchestrationService' (SKILL.md) ingests clinical notes and patient history records and interpolates them directly into system prompts without using boundary markers, XML delimiters, or input sanitization. This creates an attack surface where content in clinical records could potentially override agent instructions.
  • [EXTERNAL_DOWNLOADS]: The installation guide requires cloning a repository from a personal GitHub account ('github.com/fahad-hamid/psique-workflow-clinic') followed by 'npm install'. Users should audit the remote repository and its dependencies prior to execution.
  • [CREDENTIALS_UNSAFE]: The skill defines a configuration for accessing sensitive Argentine tax authority (AFIP) private keys and certificates from local paths ('./certificates/afip-key.pem'). While necessary for billing functionality, this requires rigorous filesystem permission management to prevent data exposure.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 11:44 PM