Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add skillmds/skillmd --skill hermes-dojogit clone --depth 1 https://github.com/skillmds/skillmdWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/skillmds/skillmd/hermes-dojo)<a href="https://agentmods.dev/skills/skillmds/skillmd/hermes-dojo"><img src="https://agentmods.dev/badge/skills/skillmds/skillmd/hermes-dojo/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/skillmds/skillmd/hermes-dojo"><img src="https://agentmods.dev/badge/skills/skillmds/skillmd/hermes-dojo.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00062 | $0.01242 |
| Opus 5.5 | $0.00025 | $0.00497 |
| Sonnet 5 | $0.00012 | $0.00248 |
| Haiku 4.5 | $0.00006 | $0.00124 |
Grade A, and why
hermes-dojo scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
This is a copy
100% identical to hermes-dojo — 1 line differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Overview
Hermes Dojo is your agent's training gym. It reads your past sessions, finds where the agent struggles, creates or improves skills to fix those weaknesses, and tracks improvement over time.
The core loop: measure → identify weakness → fix → evolve → verify → report
Commands
/dojoor/dojo analyze— Analyze recent sessions for failure patterns/dojo improve— Fix the top weaknesses (patch skills + run self-evolution)/dojo report— Show current performance metrics and improvement history/dojo history— Show learning curve over time/dojo auto— Set up overnight cron: analyze + improve + report at 6am/dojo status— Quick summary of agent health
Workflow
Step 1: Analyze
Run python3 ~/.hermes/skills/hermes-dojo/scripts/monitor.py to scan recent sessions.
This reads ~/.hermes/state.db and produces a JSON report with:
- Per-tool success/failure rates
- Error patterns (grouped by tool and error type)
- User correction signals (messages containing "no,", "wrong", "I meant", "not what I")
- Skill gap detection (repeated manual tasks with no skill)
- Session-level metrics (tool calls per session, retry patterns)
Present the results as a clear dashboard:
=== Hermes Dojo Analysis ===
Sessions analyzed: 23 (last 7 days)
Total tool calls: 156
Overall success rate: 78%
Top Weaknesses:
1. terminal_run: 73% success (12 failures) — common error: "command not found"
2. web_extract: 81% success (4 failures) — common error: "timeout"
3. No skill for: CSV parsing (requested 4 times)
User Corrections Detected: 7
- 3x wrong file path
- 2x wrong command syntax
- 2x misunderstood request
Step 2: Improve
For each identified weakness, decide the fix:
A) Existing skill fails → patch it:
- Read the current skill's SKILL.md
- Analyze the failure patterns from Step 1
- Use
skill_managewith action "patch" to add error handling, better instructions, or edge case coverage - Log the change
B) No skill exists for a recurring need → create one:
- Analyze the session patterns where this capability was needed
- Use
skill_managewith action "create" to make a new skill - Include specific instructions based on what worked in past sessions
- Log the creation
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 143 lines · 62 tokens per session scan A 084c16f3df11
hermes-dojo is a skill published in the GitHub repository skillmds/skillmd (1 stars, last pushed yesterday), licensed MIT. It adds 62 tokens to every session and 1,242 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 100% identical to hermes-dojo, differing in 1 line, and is treated as a copy.
Other skills, from other repositories
slopless
Use Slopless to review English Markdown for deterministic AI and human slop signals, including vague phrasing, formulaic prose, weak rhythm, filler, and cliches.
suiko
A Japanese writing and review workflow for checking whether text sounds uniform, translated, unnatural, or difficult to read. It supports writing and editing documents such as reports, guides, meeting notes, proposals, and emails.
poly-lint-and-format
Running poly lint / poly fmt — --fix, --format pretty|json|toon, --exclude, --config, exit codes, inline suppression, and the check → read-json → fix → re-check loop.
poly-orchestrator
Use poly as the single lint/format gate instead of invoking ruff/oxlint/rustfmt directly — one poly.toml, poly hooks install, poly migrate, CI via the Goldziher/poly setup Action.
poly-mcp
The poly MCP server — the eleven tools and which are read-only vs mutating, the paths/exclude/config/format params, async Tasks for the whole-project phase, and the isError contract.
poly-tiers-and-scope
Five mechanisms decide what actually inspects a file (crates/poly-core/src/registry.rs).