Hermes Self-Evolution
Evolve Hermes skills and prompts with DSPy + GEPA. No GPU — API calls only, ~$2-10 per run.
git clone https://github.com/NousResearch/hermes-agent-self-evolution && cd hermes-agent-self-evolution && pip install -e ".[dev]"About
Evolutionary self-improvement for Hermes Agent. Uses DSPy + GEPA (Genetic-Pareto Prompt Evolution, ICLR 2026 Oral) to mutate, evaluate, and select better versions of skills, tool descriptions, and prompts — reading execution traces to understand why variants fail. Phase 1 (SKILL.md evolution) is implemented; tool descriptions, system prompts, and code evolution are planned. Guardrails: full test suite must pass, size limits enforced, every change ships as a PR for human review — never a direct commit. Not a SKILL.md skill — a Python tool that operates on them. Install: clone the repo, `pip install -e ".[dev]"`, point HERMES_AGENT_REPO at your hermes-agent checkout. Indexed from the NousResearch/hermes-agent-self-evolution repo for discovery — Solid State has not audited this tool.
Pre-publish audit for ClawHub skills. Nine checks, one verdict: READY or FIX FIRST. Pass the security scan on the first upload.
Find and verify real research papers through a keyless index — no login, no API key. Search by topic or method, map a field's citation graph, pull the in-body passages that prove a claim, or scan GitHub for engineering prior art. Grounds every answer in papers the index actually returns — never invents titles or arXiv IDs.
Write-mode readiness audit for Claude Fable 5. Scans CLAUDE.md + skills + configs for the five documented breakage families, patches what you approve, stamps a dated FABLE-READY verdict. $49 once.