Agent Loop Engineering Kit: developer tool for Hermes Agent

Design, validate, dry-run and audit bounded Hermes Agent loop contracts before automating them

What Agent Loop Engineering Kit does

The kit installs a hermes-loop command. You write a loop spec in YAML that names the trigger, inputs and state, allowed tools, forbidden actions, risk class, deterministic checks, human gates and stop conditions. The command then validates the schema and safety gates, scores how well the loop is engineered, builds a promotion audit report, and runs a dry run that writes a run record and a readable receipt.

It is deliberately a design and audit step, not a runtime. The README states that it does not execute real agent tasks, does not create or change Hermes cron, webhook or Kanban jobs, and does not write to Hermes profiles, memory or skills. After a clean dry run, the suggested next step is one manual read-only run in Hermes with the spec as the contract. A privacy-scan command checks files for common leaks before you publish them.

Key features

hermes-loop init creates a loop spec and validate checks it against the schema and safety gates
score rates whether the loop is actually loop-engineered
audit-report builds a promotion audit report, with JSON output
dry-run writes a run record and receipt without executing the real task, and render-receipt makes it readable
privacy-scan checks a directory for common leaks before artifacts are published

When to use it

Turning a prompt like "every morning, make me a briefing" into a loop with explicit stop conditions
Reviewing a loop contract before enabling a Hermes cron job or webhook
Producing an audit record for a first manual read-only run

Who it is for: Hermes Agent users who plan repeatable or scheduled agent workflows and want them reviewed before activation.

How it fits with Hermes Agent

Built around Hermes Agent: it designs and checks loop contracts meant for Hermes cron, webhook, Kanban and GitHub automation, but it never touches a Hermes install itself.

How to install Agent Loop Engineering Kit

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

uv tool install https://github.com/AlekseiUL/agent-loop-engineering-kit.git
pipx install git+https://github.com/AlekseiUL/agent-loop-engineering-kit.git
hermes-loop --help

Requirements: Python 3.11+ and either uv or pipx

Note: Version 0.1 is distributed only as a GitHub install; PyPI publishing is listed as a later packaging step.

FAQ

What is Agent Loop Engineering Kit?

Agent Loop Engineering Kit is a Python command-line toolkit for designing, validating, scoring, dry-running and auditing Hermes Agent loop contracts. It checks the contract on paper before any real automation is switched on.

Does Agent Loop Engineering Kit work with Hermes Agent?

Yes, it is written for Hermes Agent loops. It only designs and dry-runs specs, so it does not execute agent tasks or modify Hermes cron, webhook or Kanban jobs.

How do I install Agent Loop Engineering Kit?

Run uv tool install with the repository's GitHub URL, or use pipx install git+ with the same URL, then check it with hermes-loop --help. Python 3.11 or newer is required.