verkyyi

Hermesbench — Development skill for Claude Code

Development community

Reliability-first benchmark and evaluation harness for Hermes Agent runtime configurations.

How to install Hermesbench

This entry records only its repository, not the path inside it, so there is no exact command to give. Open verkyyi/hermesbench and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Hermesbench does

Reliability-first benchmark and evaluation harness for Hermes Agent runtime configurations.

Alternatives in Development

IPolloWork — Enterprise-grade, local-first Agent Workbench for people and agent teams 4.9k ★
Petdex — A public gallery of animated pets for Codex, Claude Code, DeepSeek Harness, Hermes, OpenCode, Gemini CLI, and 4k ★
Contributing To Claude Code Configs — Thank you for your interest in contributing to Claude Code Configs 625 ★

README

HermesBench

HermesBench is a reliability-first benchmark and reusable evaluation harness for **Hermes Agent runtime configurations**.

It is not a model leaderboard. The unit under test is the whole Hermes setup: profile prompt, model/provider choice, tools, skills, memory, gateway behavior, delegation/routing, safety/refusal behavior, latency, and runtime stability.

The headline question is:

Given this Hermes configuration, does the agent reliably reach useful, truthful, stable conclusions for real user requests?

HermesBench currently targets Hermes Agent users who customize a personal agent for daily work: calendar, mail, messaging, web lookup, local context, finance, travel, reports, and optional power-user integrations.

Alpha Quick Start

HermesBench is designed to be driven through a coding agent. Start with one default scenario recipe; full bundle runs are opt-in because they take longer and cost more.

Use the HermesBench skill and run one default scenario recipe for my current Hermes configuration.

Skill: https://github.com/verkyyi/hermesbench/blob/main/agent-skills/hermesbench/SKILL.md

Follow the skill's "Run Current Hermes Configuration" workflow. Use the Python API default single-recipe path, save artifacts, and summarize the score and main findings. Do not run the full bundle unless I explicitly ask.

After a first run, open alpha feedback with the first setup issue, scoring surprise, recipe concern, or redaction/trust gap you found: [`FEEDBACK.md`](FEEDBACK.md).

What It Includes

27 bundled workflow recipes across 9 job-area categories.
Harness-driven scenarios: a use case can be one user turn or a multi-turn conversation in one isolated Hermes session.
Driver/target separation: recipes define human-facing user jobs; run configuration chooses the driver and target agent adapter.
Flat recipe categories: one visible grouping level for browsing, filtering, and optional batch runs.
**Score-only