Dynamic Skill Dojo
Hermes Dojo monitors agent sessions, identifies weak skills and missing workflows, runs targeted fixes, and reports whether the agent is improving over time, closing the loop between observability and self-evolution.
Requirements
- The Hermes Dojo skill must be present.
- Hermes session state is needed, since the performance monitor reads it to track failures.
- Optional: the Hermes self-evolution repository, for GEPA optimization.
- Optional: Hermes cron and Telegram delivery.
How to use
Treat automatic skill patches as proposed changes, and review diffs before trusting them in production workflows.
Dojo is a good fit for long-lived Hermes installations with enough session history to expose recurring patterns.
This setup guide covers installing the skill, seeding demo data, and running the analysis, improvement, and reporting commands.
The auto run proposes skill patches but should not apply them without creating a reviewable diff.
What it does: the improvement cycle runs as follows.
measure -> identify weakness -> evolve -> measure again -> report
Setup: install the project as a Hermes skill.
Clone the repository, enter its directory, and run the installer.
git clone https://github.com/Yonkoo11/hermes-dojo.git
cd hermes-dojo
./install.sh
To try it out, run the demo pipeline: change into the installed skill's scripts directory, seed demo data, and run the demo.
cd ~/.hermes/skills/hermes-dojo/scripts
python3 seed_demo_data.py --days 7
python3 demo.py --reset
Common commands cover analyzing sessions, running improvement, generating reports, viewing history, and automatic mode.
/dojo analyze
/dojo improve
/dojo report
/dojo history
/dojo auto
Prompts — analyze recent failures.
Ask Dojo to analyze the last week of sessions, rank the top three skill weaknesses, and show evidence before proposing fixes.
Use Hermes Dojo to analyze the last week of sessions. Rank the top three skill
weaknesses and show evidence before proposing fixes.
For overnight improvement, set up an auto run that analyzes failures, proposes skill patches, and delivers a report without applying patches unreviewed.
Set up a Dojo auto run that analyzes failures, proposes skill patches, and
delivers a report. Do not apply patches without creating a reviewable diff.
Source: aliaihub/awesome-hermes-usecases and its contributors. License: CC-BY-4.0. Source license notice.
Translation and editorial adaptation by ClawCases.
>Resources
>Related Use Cases
AI-Powered Earnings Tracker
Stay on top of earnings season across dozens of tech companies without manually checking multiple sources and remembering report dates for every AI/tech earnings report.
AI Video Editing via Chat
Describe your edit to an agent to trim videos, add music, subtitles, or crops — no timeline or GUI. Install two skills, then editing is done by conversation.
Screen Recording → Tutorial Video
Give Hermes a raw screen recording and it generates the tutorial script, produces the finished video on HeyGen with your AI avatar, and saves your preferences in memory for follow-up videos.