jpeetz/agent-skills

hermes-bot-team-design

Use this skill when an agent needs to design a team of specialist Hermes bots from a project specification.

查看源码

使用方法

  1. 复制安装命令,在自己的智能体环境安装前先检查源码包。
  2. 从原始文档核对所需工具、凭据和支持的宿主。
  3. 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/jpeetz/agent-skills --skill hermes-bot-team-design

开始前需了解

  • 安装不包含模型订阅、API 额度或访问文件的权限。
  • 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
仓库原始内容

按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。

Hermes Bot Team Design

Design a team of specialized Hermes bots from a project specification. Each bot gets a SOUL.md with identity, behavior rules, handoff protocols, shared Memory section, and group chat integration. MCP servers (MeMex Zero RAG, HermesVault) are wired by default.

When to use

  • A new project needs a team of specialist bots
  • You are expanding an existing fleet
  • You are migrating an OpenClaw workflow to Hermes Bot Mode

The team design workflow

1. Read the project spec

Understand the full pipeline before deciding roles. Identify: pipeline phases, which need deep reasoning (expensive model) vs high-volume/mechanical (cheap model), outputs, and handoffs.

2. Decide team size and roles

Map each phase to a bot role:

  • Discovery / Research — high-volume, cheap model (DeepSeek V4 Flash)
  • Improvement / Creation — deep reasoning, expensive model (DeepSeek V4 Pro)
  • Validation / Publishing — mechanical, cheap model (DeepSeek V4 Flash)
  • Strategy / Direction — deep reasoning, expensive model (DeepSeek V4 Pro)
  • Content / Copy — creative, expensive model (DeepSeek V4 Pro)
  • UX / Design — creative, expensive model (DeepSeek V4 Pro)
  • Frontend / Backend — implementation, cheap model (DeepSeek V4 Flash)
  • Commerce / Marketplace — mechanical, cheap model (DeepSeek V4 Flash)

3. Create a group chat directory

bash
mkdir -p ~/workspace/MeMex-Zero-RAG/wiki/agentforge/group-chat/<project-name>/

Write a README.md with: protocol (file-based messaging), team roster (bot name, role, phase, model), pipeline flow, and communication rules.

4. For each bot, create the profile

~/.hermes/profiles/<name>/
├── SOUL.md      # Identity, Memory, behavior, handoffs
├── config.yaml  # model, provider, inherit_mcp_toolsets: true

SOUL.md sections: Frontmatter (name, description, role, teammates) → Identity → Memory (shared across all profiles — MeMex + HermesVault) → Group Chat → Browser verification → Behavior → Handoffs → Agent Inbox Protocol.

config.yaml:

yaml
model:
  default: <provider>/<model>
  provider: <provider>
platforms:
  telegram:
    enabled: false  # Team bots don't need Telegram

5. Wire MCP servers (mandatory)

Set inherit_mcp_toolsets: true to inherit MeMex Zero RAG (wikisearch, wikiread, wikiquery, wikiingest, wikilint) and HermesVault (readfile, searchfiles, listdirectory, readfileheader) from global config.

6. Write roster + drop initial message

Add the roster to the group chat README. Write a message from the director introducing the team, pipeline, and first task.

7. Verify

  • ls ~/.hermes/profiles/<name>/ — SOUL.md + config.yaml exist
  • hermes gateway list — gateway loads the new profile
  • Group chat has README.md + first message

Pitfalls

  • Do NOT create a bot without a SOUL.md
  • Do NOT skip the Memory section (shared brain between all bots)
  • Do NOT use expensive models for mechanical work — Scout (discovery) and Overseer (publishing) are cheap
  • Do NOT wire Telegram to team bots — they communicate via the group chat
  • Do NOT suggest UI buttons that don't exist — if the user says "there's no such button," find an alternative
  • Verify catalog files after rebase — conflict resolution can silently drop sibling rows

References

  • references/skill-foundry-example.md — full team design example (11 bots, Skill Foundry pipeline)
来自同一仓库

更多 Skills

全部 Skills
jpeetz
社区

accessibility-compliance-audit

Comprehensive web accessibility (a11y) audit and compliance skill. Audits HTML, React, Vue, and Angular codebases against WCAG 2.2 AA standards. Activates when users ask for "accessibility audit", "a11y check", "WCAG compliance", "ADA compliance", "screen reader testing", "keyboard navigation audit", "color contrast check", "accessibility fix", or "make this accessible". Detects violations, explains impact, and generates fix-ready code. Covers semantic HTML, ARIA authoring, focus management, col

安装量
1
GitHub Stars
14
最近更新
10月21日
jpeetz
社区

agent-n-video-pipeline

Use this skill when producing a fully-local "Agent N presents a message" video on the 16GB Apple-Silicon M5 executor: Kokoro mature TTS → LTX-2.3 MLX A2V talking render with native lip sync → karaoke captions → trim → QA → X caption. Replaces the old cloud Herald pipeline (grok-imagine + volcengine) with zero cloud cost. Activates for Herald release videos, Agent N announcements, and newsletter video messages.

安装量
1
GitHub Stars
14
最近更新
10月21日
jpeetz
社区

anti-over-engineering

Use when AI agents over-engineer code beyond user requests.

安装量
1
GitHub Stars
14
最近更新
10月21日
jpeetz
社区

api-and-interface-design

Use this skill when designing APIs, module boundaries, or any public interface. Create stable, well-documented interfaces that are hard to misuse — REST endpoints, GraphQL schemas, module contracts, component props. Covers contract-first design, error semantics, validation boundaries, idempotency, and consistent naming.

安装量
1
GitHub Stars
14
最近更新
10月21日