omh-files
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user requests for filesystem operations, creating a potential surface for indirect prompt injection.
- Ingestion points: User requests defining file tasks such as list, move, or delete (SKILL.md).
- Boundary markers: Explicit instructions to use 'prepared-vs-observed boundary' and 'not-evidence boundaries' to separate guidance from actual file evidence.
- Capability inventory: Local filesystem operations including list, search, move, rename, and delete via native tools and 'omh' CLI (SKILL.md).
- Sanitization: Implements mandatory path scoping, confirmation gates for destructive actions, and checkpoint verification before execution.
Audit Metadata