Skip to content
Back to skills

Mcp Connectors

ASecurity

Use product connectors (MCP under the hood) and auth gateways without stuffing every action schema into context.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 23, 2026
ai-agentsshellgitapi

Works with

  • api
  • mcp

Security analysis

A100/100

Scanned September 23, 2026

npx -y skills add danmo-ai/danmo-work --skill mcp-connectors --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Mcp Connectors?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Mcp Connectors
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/danmo-ai-mcp-connectors/badge)](https://www.skillsdirectory.com/skills/danmo-ai-mcp-connectors)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: mcp-connectors
source: builtin
description: Use product connectors (MCP under the hood) and auth gateways without stuffing every action schema into context.
---

# Connectors

Danmo Work mounts enabled **connectors** into the Agent Loop. Each action is
exposed as a tool named `mcp_<connector>_<action>` (see system
`<mcp-tool-naming>`). Skills may cite short names; always call the full `mcp_*`
name from the tool list.

## When to use what

- Temporary / one-off HTTP → `http_request`
- Product SaaS actions (Notion, GitHub, Feishu docs) → connector actions
- Auth-heavy SaaS with many APIs → prefer a gateway connector (Composio /
  OpenConnector) instead of many raw HTTP calls
- Chat ingress (WeChat / Feishu / WeCom / QQ) is **not** a connector — it only
  delivers messages

## Gateway meta-tools (Composio-style)

If the mounted connector exposes search / connect / execute meta-actions:

1. Search for the right upstream action
2. Connect / wait for OAuth if the user is not linked yet (`ask_user` with the
   authorize link)
3. Execute; prefer batch execute when available
4. Do **not** dump secrets or tokens into replies

## Auth

If a connector returns unauthorized / 401:

1. Ask the user to open **Connectors** in the UI and complete auth (headers or
   OAuth token) for that connector.
2. Do not invent tokens or put secrets in `http_request` unless the user
   explicitly pasted a short-lived token for a one-off call.

## Permissions

- Connector actions are `external` risk — expect approval unless permission mode
  allows them; dangerous shell and external connector actions may still ask.
- Never put API keys into tool arguments when a secret header / OAuth token
  already exists on the connector.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.