client-service-agreement
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data (personal and business identification details) to populate contract templates. While this creates a potential surface for indirect prompt injection, the skill lacks sensitive capabilities such as network access or file system operations to realize an attack.
- Ingestion points: User input requested in Workflow 1 (Step 1) for party qualification (SKILL.md).
- Boundary markers: None explicitly defined for user-provided data.
- Capability inventory: No subprocess calls, network operations, or file-write capabilities detected across the instructions.
- Sanitization: No explicit sanitization or validation of the ingested strings is performed before interpolation.
- [SAFE]: The skill contains no executable code, script references, or external dependencies. All logic is contained within natural language instructions and structured workflows for legal document generation.
Audit Metadata