hermes-imports
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external Hermes workflows, which introduces an attack surface where malicious instructions embedded in those workflows could attempt to bypass the sanitization logic or influence the agent's output.
- Ingestion points: Local Hermes workflows provided as text input in SKILL.md examples and operational context.
- Boundary markers: The instructions lack specific technical delimiters to separate the untrusted input from the agent's instructions.
- Capability inventory: The skill focuses on generating sanitized text and summaries; no direct file system or network capabilities are invoked in the instruction set.
- Sanitization: The skill incorporates a robust Sanitization Checklist and explicit Import Rules to filter sensitive information, serving as a primary defense mechanism.
Audit Metadata