hermes-skill-audit

Pass

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing untrusted content from other Hermes skills, creating an indirect prompt injection surface.\n
  • Ingestion points: The skill enumerates and parses SKILL.md files, scripts, and metadata from global, tap-installed, and profile-local directories.\n
  • Boundary markers: A dedicated 'Untrusted Content Boundary' in SKILL.md and references/safety.md explicitly instructs the agent to treat audited content as evidence rather than instructions.\n
  • Capability inventory: The tool is restricted to read-only file system operations; it lacks network access, shell execution, or file-write capabilities in its default configuration.\n
  • Sanitization: The agent is instructed to use minimal quoting and specifically flag suspected injection attempts as findings rather than obeying them.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 25, 2026, 07:07 PM
Security Audit — agent-trust-hub — hermes-skill-audit