The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a custom CLI suite (omh) to perform memory management tasks, including status checks, data capture, and incident diagnosis. These commands are integral to the workflow and are part of the vendor's specialized toolset provided for the Hermes environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and review content from USER.md and MEMORY.md. Since these files contain user-supplied data, they represent a potential attack surface for indirect prompt injection where malicious instructions could be embedded in memory entries.
Ingestion points: USER.md and MEMORY.md (read via tool calls and shell enumeration in SKILL.md and references/source-recovery.md).
Boundary markers: The skill instructions specify the use of 'not-evidence boundaries' and explicitly tag context as not_omh_reviewed to distinguish between provided data and reviewed claims.
Capability inventory: The skill has access to memory file write operations (via the omh_memory tool) and extensive shell execution capabilities through the omh CLI suite.
Sanitization: The protocol mandates quoting only observed claims and strictly separating prepared guidance from observed evidence to prevent the agent from inventing provenance or misinterpreting data as instructions.

omh-memory-sync