
from hermes-dojo83
Automated self-improvement system that analyzes agent sessions, identifies recurring failures or missing skills, and proposes or patches skills to improve perfo
Hermes Dojo analyzes an agent's past sessions to detect failure patterns, missing capabilities, and recurring user corrections. It automates a fix loop: measure weaknesses, create or patch skills, run targeted self-evolution, and verify improvements—then generates concise reports.
Use Hermes Dojo when you want the agent to autonomously discover systematic failures across sessions (e.g., repeated command errors, parsing failures, or missing task skills), run corrective skill patches or create new skills, and track progress over time. It's intended for overnight or scheduled runs to continuously raise the agent's baseline performance.
Best suited for agents with terminal and local-session access (agents that can run Python/Bash and read session logs). Works with Claude/Codex-style agents and custom Hermes agent toolsets that support skill_manage and session inspection.
This skill has not been reviewed by our automated audit pipeline yet.