Refine Cycle: self-improvement plugin for Hermes Agent Refine Cycle, a self-improvement plugin for Nous Research's Hermes Agent, automatically detects recurring errors across sessions and applies up to three reversible changes per day to memory, skills, or prompts, with all mutations journaled and credentials redacted. The plugin adapts Prime Intellect's Prime Agent /refine concept, requires a full installer that connects to the session's AI model and expands long-term memory, and includes a rollback command via `python install.py --rollback`. Your agent keeps repeating the same mistake. This makes it stop. Refine Cycle looks across recent sessions, finds those repeating problems, and saves one small lesson when the evidence is strong enough. Later, it checks whether the same problem came back. Cross-session by design. Hermes can learn from the conversation in front of it, but some problems return across different sessions: the same failed command, the same wrong assumption, the same workaround you have to explain twice. Underneath: errors are fingerprinted into comparable shapes, recurrence is counted within and across sessions , and every mutation is journaled before it runs. It adapts the /refine concept from Prime Intellect's Prime Agent https://www.primeintellect.ai/blog/prime-agent Continual Harness to the Hermes plugin system. 1. Notice what keeps going wrong. One bad result may be noise. A problem seen in two sessions or five times is a pattern worth examining. 2. Save the smallest useful lesson. It can add a short memory, create or improve a reusable skill, or add a focused note for future turns. 3. Check the result. It watches later sessions and reports whether the lesson appears to be working, unused, unreliable, or too new to judge. - It makes no more than three changes per day. - Every change is recorded. When it can be safely undone, it gives you one command to reverse it. - It never rewrites Hermes's base instructions or deletes your skills. - API keys and other credentials are removed before conversation evidence is sent to the model. - If the evidence, model reply, or Hermes state is unclear, it stops instead of pretending that a lesson was applied. Refine Cycle does more than report problems: it can change what Hermes remembers. The full installer connects Refine Cycle to the AI model already serving your Hermes session and increases the space available for long-term memory. When the plugin starts, it also attempts to turn off Hermes's manual memory and skill approval queues so lessons do not remain pending forever. Those changes are disclosed, backed up where applicable, and reversible through python install.py --rollback . See Installation installation for the exact files, commands, and host-version checks before you run it. Technical documentation starts here. The sections below describe the signal and application gates, journal states, host patch, privacy boundaries, rollback, and test evidence. trajectory state.db → scrub → fingerprint + aggregate → signal gate ├→ reviewer decline → journaled no op └→ proposal → guardrails + prepare → apply → finalized outcome → usefulness ledger | Stage | What happens | |---|---| | 1. Collect evidence | Reads the last N messages of the selected session from