Webhook Token Security (Zero‑Exposure Edition)

Secure webhook token management using MGC Blackbox. Supports DingTalk, WeCom, Feishu, Telegram, Slack and more. Store webhook tokens locally in encrypted form, retrieve at runtime without exposing to AI models.

Overview

Webhook Token Security is a documentation skill that teaches how to manage webhook tokens securely using MGC Blackbox. Supports multiple platforms including DingTalk, WeCom (Enterprise WeChat), Feishu (Lark), Telegram, and Slack. It enables AI agents to send notifications without ever exposing webhook tokens to the AI model.

This skill contains no executable code and is safe for automatic approval.

---

What This Skill Enables

After reading this documentation, an AI agent will understand how to:

• Store webhook tokens (DingTalk, WeCom, Feishu, Telegram, Slack, etc.) securely in MGC Blackbox
• Retrieve tokens at runtime without AI seeing plaintext
• Send notifications through local scripts (Mode A) or via mgc_run (Mode B, full zero-exposure)
• Pass message content to scripts via the ext02 parameter
• Handle platform-specific differences
• Rotate tokens without code changes

---

Prerequisites

1. Install MGC Blackbox: pip install mgc-blackbox (recommended 1.4.9+)
2. Start MGC service: mgc (WebUI: http://127.0.0.1:57218, API: http://127.0.0.1:57219)
3. Use MCP tools (mgc_save, mgc_run, mgc_list, mgc_open_webui, mgc_seal) for token & script management

Important: mgc_get returns plaintext and is for human/debug use only — it breaks zero-exposure. Prefer mgc_run for AI-driven sending.

Sandbox mode (Trae Work / Workbuddy): After installing MGC, open the WebUI to view and install the main MGC skill documentation. Run mgc --status to check status and sandbox mode.

---

Supported Platforms

Platform	Token Type	Storage Format	API Endpoint
DingTalk	access_token + secret	JSON	https://oapi.dingtalk.com/robot/send
WeCom	webhook key	Plain text	https://qyapi.weixin.qq.com/cgi-bin/webhook/send
Feishu	webhook_url	Plain text	Custom webhook URL
Telegram	bot_token	Plain text	https://api.telegram.org/bot{token}/sendMessage
Slack	webhook_url / bot_token	JSON	Incoming Webhook or Web API

---

Platform-Specific Storage

DingTalk

Requires both access_token and secret for signature verification.

json

{
  "access_token": "your_access_token",
  "secret": "your_secret",
  "webhook": "https://oapi.dingtalk.com/robot/send?access_token=xxx"
}

Storage key: info_type=config, info_owner=dingtalk_myapp

WeCom (Enterprise WeChat)

Requires only the webhook key from the custom robot configuration.

json

{
  "webhook_key": "your_webhook_key",
  "webhook_url": "https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=xxx"
}

Storage key: info_type=config, info_owner=wecom_myapp

Feishu (Lark)

Requires the webhook URL from the custom bot configuration.

json

{
  "webhook_url": "https://open.feishu.cn/open-apis/bot/v2/hook/xxx"
}

Storage key: info_type=config, info_owner=feishu_myapp

Telegram

Requires bot_token and optionally chat_id.

json

{
  "bot_token": "your_bot_token",
  "chat_id": "your_chat_id"
}

Storage key: info_type=config, info_owner=telegram_mybot

Slack

Can use either incoming webhook URL or bot token.

json

{
  "webhook_url": "https://hooks.slack.com/services/xxx",
  "bot_token": "xoxb-xxx",
  "channel": "#my-channel"
}

Storage key: info_type=config, info_owner=slack_myapp

---

Storing Webhook Tokens

Step 1: Prepare Token File

Create a JSON file containing your webhook token details (see Platform-Specific Storage above).

Step 2: Store in MGC

Important: Tokens should be stored by humans via WebUI to avoid AI directly handling sensitive values. AI may call mgc_open_webui to open the page for the user.

Recommended: WebUI (for human operators)

1. Open: http://127.0.0.1:57218
2. Navigate to Save page
3. Enter info_type: config, info_owner: your_webhook_name
4. Enter token content
5. Click Save

Alternative: MCP (for AI agents, when user-authorized)

• Store token: mgc_save(info_type="config", info_owner="...", content=...)
• List stored tokens (no plaintext): mgc_list(info_type="config")

---

Two Execution Modes

Mode A: Local Script + MGC-stored Config (token zero-exposure)

Script runs locally and reads the token from MGC via HTTP API at runtime. The token never appears in code, but the script source is visible to whoever runs it.

1. User stores token via WebUI (info_type=config)
2. Local script reads token from MGC API at runtime
3. Script formats message and sends HTTP POST
4. Returns non-sensitive result only

Suitable for: one-off or debug tasks where you control the host.

Mode B: Script in MGC + mgc_run (full zero-exposure) — recommended

Script is stored (and optionally sealed) inside MGC. AI calls mgc_run to execute; MGC returns only the execution result. AI never sees the token, the script source, or stdout.

1. Store token: mgc_save(info_type="config", info_owner="dingtalk_myapp", content=...)
2. Store sending script: mgc_save(info_type="script", info_owner="webhook_send_dingtalk_v1", ext01="python", content=...)
3. (Optional) Seal script: mgc_seal(info_owner="webhook_send_dingtalk_v1") — sealed scripts can only run inside MGC, cannot be decrypted
4. AI executes: mgc_run(info_owner="webhook_send_dingtalk_v1", ext02=json.dumps({"message": "deploy ok"}))
5. MGC returns execution result only

Suitable for: production, multi-agent collaboration, any case where AI must not touch the token.

Note: mgc_run returns only the execution result, not script stdout. For sending tasks, have the script return a status JSON as the result; if detailed output is needed, write to a file and return the path.

ext02 Parameter (passing message content)

ext02 carries runtime params to the script. It MUST be a JSON string (use json.dumps()). Some MCP clients mis-serialize dict values and return HTTP 422.

python

import json
ext02 = json.dumps({"title": "Alert", "message": "Deploy succeeded"})
result = mgc_run(info_owner="webhook_send_dingtalk_v1", ext02=ext02)

The script reads ext02 from its input, fetches the token from MGC internally, and sends. Token and script source stay inside MGC.

---

MGC Blackbox API Reference

Service Endpoint

• Base URL: http://127.0.0.1:57219
• Token File: ~/.mgc/database/mgc_black_box/.mgc_token
• Token: String token read from token file, required for all API calls

Get Token API

Endpoint: /api/mgc/sensitive/get Method: POST Headers:

• X-MGC-Token: (string token read from token file)
• Content-Type: application/json

Body fields:

• info_type: "config"
• info_owner: your chosen identifier

Response fields:

• code: status code
• data.content: JSON string containing stored token

Save Token API

Endpoint: /api/mgc/sensitive/save Method: POST Headers: same as above

Body fields:

• info_type: "config"
• info_owner: your identifier
• content: JSON string of token

Run Script API (mgc_run, since 1.4.7)

Endpoint: /api/mgc/sensitive/run Method: POST Headers: same as above

Body fields:

• info_type: "script"
• info_owner: script name
• ext02: JSON string of runtime params

Response: execution result only (non-blocking since 1.4.5, may return PID immediately). mgc_get action=run is retained for backward compatibility.

---

Security Best Practices

1. Never embed tokens in code
2. Use MGC for token storage
3. Retrieve tokens at runtime only
4. Never log or print tokens
5. Rotate tokens regularly
6. Use separate tokens per platform/per bot
7. Limit webhook permissions (send-only where possible)

---

Use Cases

• Deployment notifications
• CI/CD pipeline alerts
• System monitoring alerts
• Team collaboration bots
• Automated workflow triggers

---

Learn More About MGC Blackbox

Want to learn more about MGC Blackbox?

• Visit: https://github.com/zkeviny/MGC-Blackbox
• Report issues: https://github.com/zkeviny/MGC-Blackbox/issues
• Contact: [e-mail]

---

License

MIT