client-intake

Pass

Audited by Gen Agent Trust Hub on Jul 12, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by processing untrusted data from client interviews and using it to generate structured outputs and commands.
  • Ingestion points: Untrusted data enters the system through client narratives and responses to intake questions described in SKILL.md (Step 1: Practice Area Routing and Step 2: Practice-Area Specific Intake).
  • Boundary markers: There are no explicit boundary markers or instructions to the agent to ignore or escape instructions embedded within the user-provided legal narratives.
  • Capability inventory: The skill produces executable-style command blocks (e.g., /legal-clinic:deadlines --add) as a 'mandatory deliverable'. If these outputs are processed by other automated tools or the agent's command line, they provide a vector for command argument injection.
  • Sanitization: The instructions do not define any sanitization, filtering, or validation for the data interpolated into critical fields such as case, description, or source within the generated deadline commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 12, 2026, 01:02 AM
Security Audit — agent-trust-hub — client-intake