← 返回 Skills 市场
davedean

Agent Contact Card

作者 davedean · GitHub ↗ · v0.1.0
cross-platform ⚠ suspicious
2086
总下载
0
收藏
2
当前安装
1
版本数
在 OpenClaw 中安装
/install agent-contact-card
功能描述
Discover and create Agent Contact Cards - a vCard-like format for AI agents. Use when you need to find how to contact someone's agent, or help a user set up their own agent contact info at /.well-known/agent-card.
使用说明 (SKILL.md)

Agent Contact Card

A simple format for publishing how AI agents can be contacted. Like a vCard, but for agents.

When to Use This Skill

  • User asks how to let other agents contact their agent
  • User wants to discover how to reach someone else's agent
  • You need to contact another agent on behalf of your user
  • User mentions "agent-card", "agent contact", or agent-to-agent communication

Quick Reference

Discovering an Agent Contact Card

Try fetching /.well-known/agent-card on their domain:

https://example.com/.well-known/agent-card

The file is markdown with YAML frontmatter. Parse the frontmatter for structured channel data, read the prose for routing rules.

Creating an Agent Contact Card

Create a markdown file with YAML frontmatter:

---
version: "1"
human_contact: "+1 555 123 4567"
channels:
  email: "[email protected]"
  discord: "my-agent#1234"
  webhook:
    url: "https://example.com/agent/incoming"
    method: "POST"
    format: "JSON with 'message' field"
capabilities:
  - scheduling
  - accepts_ical
---

# My Agent

If you're a human, call the number above.

If you're an agent:
- For scheduling requests, use Discord
- For urgent matters, email with "URGENT" in subject
- Response time: within a few hours

Host this at /.well-known/agent-card on the user's domain.

Format Details

Required Fields

Field Description
version Spec version. Currently "1"

Recommended Fields

Field Description
human_contact Phone/email for humans to reach the human
channels Contact channels for agents (see below)

Optional Fields

Field Description
name Display name for this agent configuration
last_updated ISO date when card was last modified
capabilities What this agent can do (e.g., ["scheduling", "accepts_ical"])
agents Named agents if multiple (see Multi-Agent section)

Channels

Channel names are freeform. Common ones:

  • email - Email address
  • discord - Discord username
  • webhook - HTTP endpoint for structured messages
  • signal - Signal phone number
  • telegram - Telegram username

For webhooks, provide details:

channels:
  webhook:
    url: "https://example.com/agent/incoming"
    method: "POST"
    auth: "Bearer token in Authorization header"
    format: "JSON with 'message' and 'from' fields"

Multi-Agent Setups

List multiple specialized agents:

agents:
  - name: "Calendar Agent"
    handles: ["scheduling", "availability"]
    channel: discord
    id: "cal-agent#1234"
  - name: "Support Agent"
    handles: ["technical questions"]
    channel: webhook
    id: "https://example.com/support"

The markdown body should explain routing between them.

Privacy Tiers

Different URLs for different access levels:

Tier URL Pattern Access
Public /.well-known/agent-card Anyone
Named /.well-known/agent-card/{name} Know the name
Private /{random-uuid}/agent-card.md Shared URL only

Each tier can expose different channels and capabilities.

Discovery Methods

  1. Well-known URL: Check https://domain/.well-known/agent-card
  2. vCard extension: Look for X-AGENT-CARD field in contact cards
  3. Ask the human: Request the URL directly

Reading an Agent Card

When you fetch an agent card:

  1. Parse YAML frontmatter for structured data
  2. Read markdown body for natural language routing rules
  3. Choose appropriate channel based on your purpose
  4. Follow any authentication requirements mentioned

Test It

Here's a live demo you can test:

https://city-services-api.dave-dean.workers.dev/.well-known/agent-card

This is a fictional "City of Millbrook" tip line. Fetch the card, then try POSTing to the webhook endpoint. Your experience may vary depending on what you say.

Full Specification

See references/SPEC.md for the complete specification.

Examples

See references/EXAMPLES.md for more complete examples.

安全使用建议
This skill is a harmless format/spec for publishing agent contact information. Before installing or using it, be aware that: (1) following the instructions will cause the agent to make outbound network requests (fetching /.well-known/agent-card on domains and optionally POSTing to webhook URLs); avoid having the agent send sensitive data to unknown endpoints. (2) If you host an agent-card, do not place secrets (API tokens, private keys, passwords) directly in the card; use documented auth mechanisms (signed messages, short-lived tokens) and advertise how to authenticate rather than embedding secrets. (3) When testing webhooks, use controlled/test endpoints you trust. Otherwise, the skill is internally consistent with its stated purpose.
功能分析
Type: OpenClaw Skill Name: agent-contact-card Version: 0.1.0 The skill is designed to facilitate agent-to-agent communication, which inherently involves making outbound HTTP requests to external domains. The `SKILL.md` and `references/EXAMPLES.md` files explicitly instruct the agent to perform HTTP POST requests to an external domain (`https://city-services-api.dave-dean.workers.dev`) as part of a 'test' or demonstration. While these network calls are aligned with the skill's stated purpose, the ability to make arbitrary outbound network calls and send user-derived input to external endpoints represents a significant risk, even without clear evidence of intentional malicious data exfiltration or other harmful actions.
能力评估
Purpose & Capability
Name/description (agent contact discovery/creation) match the contents of SKILL.md and the supplemental spec/examples. The skill is purely a format/spec and discovery guide and does not request unrelated binaries, environment variables, or credentials.
Instruction Scope
The instructions tell agents to fetch /.well-known/agent-card, parse YAML frontmatter, read prose for routing, and (for testing) POST to webhook endpoints discovered in cards. This is expected for a contact/discovery spec, but note that following these instructions causes network requests and could send content to external webhooks if executed.
Install Mechanism
No install spec and no code files — the skill is instruction-only, so nothing is written to disk or installed during activation.
Credentials
The skill declares no required environment variables, credentials, or config paths. The spec mentions how cards may reference auth tokens or public keys, but the skill itself does not demand secrets from the environment.
Persistence & Privilege
The skill does not request persistent privileges (always:false). It does not attempt to modify other skills or system configs and does not require being force-included.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agent-contact-card
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agent-contact-card 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v0.1.0
- Initial release of agent-contact-card skill. - Enables discovery and creation of Agent Contact Cards—a vCard-like format for AI agents. - Provides clear instructions for publishing and finding agent contact information via /.well-known/agent-card. - Supports multiple communication channels and privacy tiers. - Includes specification and usage examples.
元数据
Slug agent-contact-card
版本 0.1.0
许可证
累计安装 2
当前安装数 2
历史版本数 1
常见问题

Agent Contact Card 是什么?

Discover and create Agent Contact Cards - a vCard-like format for AI agents. Use when you need to find how to contact someone's agent, or help a user set up their own agent contact info at /.well-known/agent-card. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 2086 次。

如何安装 Agent Contact Card?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-contact-card」即可一键安装,无需额外配置。

Agent Contact Card 是免费的吗?

是的,Agent Contact Card 完全免费(开源免费),可自由下载、安装和使用。

Agent Contact Card 支持哪些平台?

Agent Contact Card 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Agent Contact Card?

由 davedean(@davedean)开发并维护,当前版本 v0.1.0。

💬 留言讨论