Skills
Navigate
skills/rlaope/oh-my-hermes/omh-jev-route/Gen Agent Trust Hub

omh-jev-route

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user messages and structured routing data to determine workflow fitness, which could be exploited to influence agent behavior.
  • Ingestion points: Processes user input via the state parameter and the route_question block as specified in SKILL.md (Procedure step 2) and references/full-contract.md (Required inputs).
  • Boundary markers: Mandates a verbal notification to the user ("message leaves the machine") before data is sent, though no technical delimiters are enforced for the content itself.
  • Capability inventory: Employs the omh_jev_ask tool to perform external requests and omh_route_answer to write transaction logs to a local file (asks.jsonl).
  • Sanitization: Contains explicit logic to detect and remove credential_like_content from inputs before processing to prevent accidental secret exposure.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 12:36 AM