wysie/hermes-omni-plugin — reverse-engineered prompt

Build me a Python plugin for Hermes Agent that connects it to the local OMNI Semantic Signal Engine. I want Hermes terminal output to be automatically cleaned up and summarized before it gets sent into the agent context, with a clear label saying the output was distilled.

Please include tools for compressing any text through OMNI, getting the original raw output back from OMNI rewind when the summary is too short, checking token and cost savings, running OMNI diagnostics, and showing plugin status. Add a simple /omni slash command for status, stats, doctor, and config info.

Keep it safe by default. It should not run shell commands unless the user explicitly turns that on in the config. It should sanitize risky environment variables, use the local omni binary only, preserve raw output if OMNI fails, and have a JSON config under the Hermes plugin data folder.

Make it installable into the Hermes Python environment, with tests and clear README instructions. Look up current Hermes plugin docs online if needed.