The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various omh CLI commands, such as omh list, omh doctor, and omh status, to retrieve information about the local installation. These commands are part of the OMH vendor toolkit. The skill specifically instructs the agent to disclose the side effects of diagnostic commands like omh doctor before execution.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, including GitHub repository files and local tool outputs. It implements defensive measures by instructing the agent to treat external content as evidence rather than instruction sources and to ignore any embedded prompts or commands within retrieved documents.
Ingestion points: Files retrieved from the rlaope/oh-my-hermes GitHub repository and the output of omh CLI commands.
Boundary markers: Instructions are present to clearly label and separate public product facts from local installation facts and to cite specific versions or commits.
Capability inventory: The skill uses subprocess calls to run the omh CLI suite and performs file system reads within the OMH home directory (~/.omh).
Sanitization: The skill contains explicit rules forbidding the reading or printing of credentials, tokens, or .env files, and it mandates ignoring instructions found within fetched documentation.

omh-docs