omh-memory-new
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied information to create a
memory_new_candidate/v1artifact for future sessions. This creates a surface where malicious instructions embedded in captured data could influence the agent in subsequent turns. - Ingestion points: User prompts and context facts are ingested to generate metadata candidates as defined in SKILL.md.
- Boundary markers: The skill explicitly references "evidence boundaries" and "not-evidence boundaries" to distinguish between different types of context.
- Capability inventory: The skill utilizes the
omhCLI tool to record runtime state. - Sanitization: Instructions include a specific "Refuse" policy targeting prompt-injection-shaped instructions, secrets, raw logs, and transcripts.
- [COMMAND_EXECUTION]: The skill uses a project-specific CLI utility
omh(e.g.,omh runtime record) to track workflow phase transitions and record delegation results. This is a functional component of the "Oh My Hermes" workflow orchestration.
Audit Metadata