English | 简体中文

hermes-kit

Plugins, skills, and field notes for running Hermes Agent day to day, plus docs for a patched Hermes fork.

Quick start

I want to	Run
Install a Hermes plugin	hermes plugins install Adonis0123/hermes-kit/plugins/hermes-lock-screen && hermes plugins enable hermes-lock-screen
Register this repo as a Hermes skill source (tap), then hermes skills install <name>	hermes skills tap add Adonis0123/hermes-kit
Install one skill in Hermes	hermes skills install Adonis0123/hermes-kit/skills/cursor-cli
Use the skills in Claude Code	/plugin marketplace add Adonis0123/hermes-kit then /plugin install hermes-kit@hermes-kit
Use a skill in Codex or other agents	npx skills add adonis0123/hermes-kit --skill cursor-cli
Run my patched Hermes fork	See docs/hermes-fork.md

More options (pinning, updating, uninstalling): docs/install.md.

Contents

Plugins

Plugin	What it does
hermes-lock-screen	Owner-only Feishu DM command and tool that locks a macOS screen and verifies the lock. Design: docs/hermes-lock-screen-design.md.

Skills

Skills are flat: skills/<name>/SKILL.md.

Skill	What it does
hermes-upstream-check	Check the Hermes Agent upstream repository for new commits and releases.
hermes-daily-review	Review the previous day's Hermes sessions before editing skills.
ai-hotspot-digest	Build a daily AI digest (new skills, GitHub projects, news) for a group chat.
feishu-chat-style-review	Review how the agent replies in Feishu groups, using chat history.
feishu-ui-feedback	Turn Feishu messages with UI screenshots into concrete UI fixes.
desktop-browser-chrome	Open URLs in the user's real Chrome instead of a headless browser.
cursor-cli	Run the Cursor CLI (cursor-agent) from Hermes: print mode, streaming, model choice.
writing-with-cursor	Draft or polish multi-paragraph prose and docs through Cursor.
convert-documents-to-markdown	Convert Word, PowerPoint, Excel, and PDF files to Markdown.
figma-copy-proofread	Compare a Figma frame's copy with the live page and list the mismatches.

Docs

Doc	Topic
Install	Every install path per host, pinning, updating, uninstalling
Hermes fork	What the fork is, who should use it, how to switch
Feishu customizations	What each fork patch changes for Feishu users
Memory method	A write gate for MEMORY.md, USER.md, and SOUL.md, with templates
ADRs	Why this repo is exported, and why the fork rebases
Glossary	Terms used across these docs

Guides

Method notes from running Hermes day to day.

Guide	Topic
Memory provider adoption gate	Hot, warm, and cold memory; when an external memory provider is worth it
Feishu gateway black-box checklist	Six manual checks with message-id evidence after a config change or upgrade
Quiet delivery in Feishu groups	One message per task on the main timeline, work inside topics, noise off
Remote lock screen design	How hermes-lock-screen authorizes, runs, and verifies a Mac lock
Cron digest pipeline pattern	A thin daily-digest pipeline: filter, dedup, single-channel delivery
Upgrade-safe customization	What survives hermes update, where customizations belong, overlay tests

Agent-agnostic skills and a list of recommended third-party skills live in adonis-skills.

How it works

flowchart LR
  subgraph private["Private source repos"]
    cfg["Private config repo"]
    shared["Private shared-skills repo"]
  end
  cfg --> export["Export script<br/>allowlist, local-* excluded"]
  shared --> export
  export --> scan["Privacy scan<br/>gitleaks + denylist"]
  scan -->|"one fresh commit"| kit["hermes-kit"]
  scan -->|"one fresh commit"| as["adonis-skills"]
  up["NousResearch/hermes-agent"] -->|"rebase overlay"| fork["Adonis0123/hermes-agent"]
  kit --> hermes["Hermes"]
  kit --> cc["Claude Code"]
  as --> other["Codex and others<br/>via npx skills"]
  fork -->|"hermes update"| hermes

Loading

1. Private repos are the single source of truth. Nothing is edited here by hand.
2. An export script copies only allowlisted paths. Files named local-* hold personal or company specifics and never leave the private repo.
3. A privacy scan runs before each export, and again in this repo's CI (privacy-scan.yml).
4. Each export lands as one new commit, so no private history is published.
5. The Hermes fork is a separate flow: upstream plus a small overlay, re-applied by rebase on each upstream release.

The reasoning is in ADR 0001 and ADR 0002.

For AI agents

• AGENTS.md: install commands per host, repo layout, naming rules, validation commands.
• llms.txt: a short index of this repo for LLMs.
• skills/ and plugins/ are generated. Open an issue or PR here and the change will be made in the source.

Related

• adonis-skills: agent-agnostic skills and recommended third-party skills.
• Adonis0123/hermes-agent: the patched Hermes fork.
• NousResearch/hermes-agent: upstream Hermes Agent.

License

MIT