The Agent Skills Directory

[SAFE]: The skill serves as an instructional guide for the 'oh-my-hermes' platform to assess external connectors (APIs, plugins, memory providers). Its primary focus is security and observability, instructing the agent to verify trial results rather than relying on static documentation.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting data about untrusted external components. It provides robust mitigation strategies by enforcing 'prepared-vs-observed' boundaries, requiring explicit markers for missing or risky data, and preventing the agent from claiming successful execution without observed evidence. This structural approach minimizes the risk of the agent being misled by instructions embedded in external connector documentation.
[COMMAND_EXECUTION]: The skill references platform-specific commands such as hermes plugins install and omh runtime record, as well as standard package managers like npx, uvx, and pipx. These references are descriptive and intended to explain the platform's security posture (e.g., malware scanning) rather than instructing the agent to execute arbitrary or dangerous shell commands.

omh-external-connector-readiness