omh-people-ops
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-provided data to generate recruitment artifacts, which presents a surface for indirect instructions to influence agent behavior.
- Ingestion points: The skill ingests untrusted data through required inputs such as 'role or people-process outcome', 'available evidence' (which includes candidate interview feedback), and 'policy constraints'.
- Boundary markers: There are no explicit technical delimiters (e.g., XML tags or clear 'ignore' instructions) used to isolate these inputs from the system instructions when generating outputs like scorecards or briefs.
- Capability inventory: The skill uses
omh runtime recordfor status tracking andskill_viewfor loading shared context. It references tools likeomh-apps(for ATS/Calendar integration) but contains explicit instructions prohibiting their use within this specific workflow, reducing the risk of unauthorized execution. - Sanitization: No specific sanitization or validation logic is defined for the content extracted from 'available evidence' or other user inputs before they are synthesized into final documents.
Audit Metadata