What is hermes-bench?

A workflow for running the hermes-agentic-bench test battery against one or more local language models. The tests can use a simulated OpenAI-compatible endpoint, the real Hermes command-line tool, or both, and produce a comparison report.

Use it to install the benchmark's dependencies, run selected simulated or Hermes tests, and compare the resulting model performance.

It handles the setup, command options, and result collection needed to compare models on agent tasks.

How vcruz305/hermes-agentic-bench configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to vcruz305/hermes-agentic-bench. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/vcruz305/hermes-agentic-bench/main/.claude/skills/hermes-bench/SKILL.md

Clone the repo

git clone --depth 1 https://github.com/vcruz305/hermes-agentic-bench

Made for: Claude Code.

Wrote 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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/vcruz305/hermes-agentic-bench/hermes-bench/github.svg)](https://agentmods.dev/skills/vcruz305/hermes-agentic-bench/hermes-bench)

Your own site

<a href="https://agentmods.dev/skills/vcruz305/hermes-agentic-bench/hermes-bench"><img src="https://agentmods.dev/badge/skills/vcruz305/hermes-agentic-bench/hermes-bench/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/vcruz305/hermes-agentic-bench/hermes-bench"><img src="https://agentmods.dev/badge/skills/vcruz305/hermes-agentic-bench/hermes-bench.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 48 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 619 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

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	$0.00048	$0.00619
Opus 5	$0.00024	$0.00309
Sonnet 5	$0.00010	$0.00124
Haiku 4.5	$0.00005	$0.00062

Measured 10d ago against content hash ceae80f985e5, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

hermes-bench 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 10d 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.

.claude/skills/hermes-bench/SKILL.md · 49 lines

How it starts

The opening of the file, as written. The whole thing — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.

hermes-bench

Runs this repo's test batteries end-to-end and hands back a comparison report, without the user having to remember flags or file names.

Trigger

When the user runs /hermes-bench, or asks to benchmark/compare model(s) against this repo's agentic tests.

Workflow

Confirm location. Check for README.md and simulated_battery.py in the current directory. If missing, ask whether to clone https://github.com/vcruz305/hermes-agentic-bench here or run the skill somewhere else.
Install dependencies. pip install -r requirements.txt.
Read README.md, especially "Known limitations." The file-toolset tests are not reliably sandboxed to a scratch directory — keep the destructive-request test (--enable-destructive) off unless the user explicitly asks for it in this conversation. Don't infer consent from a prior run.
Collect targets. For each model the user wants tested, ask:
- a short label (used to name result files)
- which battery: simulated, real Hermes CLI, or both
- simulated: OpenAI-compatible --base-url, --api-key, --model
- Hermes: --provider and --model exactly as registered in the user's Hermes config.yaml — ask rather than guess, or read the file if the user points to it
Run the batteries, one model at a time:
- python simulated_battery.py --base-url <url> --api-key <key> --model <model> --output results_<label>.json
- python hermes_native_battery.py --provider <provider> --model <model> --output results_<label>_hermes.json
- Hermes-native runs can legitimately take minutes per test (900-1800s stream timeouts are expected for local providers) — don't treat a slow-but-running process as hung.
Generate the report once every requested model has finished: python generate_report.py results_*.json --output comparison.md
Report back. Show comparison.md and call out anything that looks off before the user reads the raw numbers — a test marked "not run", a large elapsed-time outlier versus the other models, a non-zero return code.

Read the full file on GitHub · 49 lines

Changes

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.

10d ago First seen · 49 lines · 48 tokens per session scan A ceae80f985e5

Subscribe to this mod's changes

hermes-bench is a skill published in the GitHub repository vcruz305/hermes-agentic-bench (14 stars, last pushed 27d ago), licensed MIT. It adds 48 tokens to every session and 619 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Building agents Testing

research-engineer

An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.

davila7/claude-code-templates · 43 tokens

tika-eval-compare

Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".

apache/tika · 50 tokens

neuron-evaluation-engineer

Create and run AI evaluations with datasets, assertions, and output drivers in Neuron AI. Use this skill whenever the user mentions evaluation, testing AI systems, creating evaluators, dataset-driven testing, assertion-based validation, or wants to measure AI system performance. Also trigger for tasks involving…

neuron-core/neuron-ai · 77 tokens

hermes-agentic-bench: Skill for Claude Code