WorkBuddy's Skill is more like a set of reusable task practices, while MCP is more like an interface connecting external tools and data sources. To have AI process files, generate content, or execute scripts in fixed steps, first find a skill; If you want to connect WeCom's bot, database, or business system to make the model truly call external services, then look to MCP. The two can work together, not replace each other.
Let's first use a table to clarify
| Comparison items | Skill | MCP |
|---|---|---|
| Main function | Encapsulated instructions, scripts, and workflows | External context and tools are open to the model |
| Common entrance | Skills marketplace, upload skill packs, find or create skills | MCP server configuration in sidebar plugins |
| Typical scene | Document processing, web search, and fixed output workflows | Notify group chats, query business data, and call third-party interfaces |
| Key points for troubleshooting | Whether installation, activation, and permissions are appropriate | Configure formatting, run commands, credentials, and connection status |
Why can't I call the skill even after installing it?
First, go to the "Installed" list to confirm the switch. Disabling Skill will only pause the model call and will not delete the skill file; Reopening it will restore functionality, and the activation status will be synchronized across multiple devices with the account. The task description should also clearly state the goal and deliverable; otherwise, the model may not decide which Skill to call. When there are too many skills, only enable the currently needed items to reduce false triggers.
MCP configuration should be at the user level or the project level
Notification services used across projects are suitable for user-level configuration; If only a data source serves a specific client or project, it should be placed under the project-level configuration directory to avoid misuse by other tasks. After saving, first check the MCP Server status. Green indicates successful connection, red indicates priority checks for JSON parentheses and quotes, whether the startup command exists, whether credentials are valid, and whether the network is reachable.
How to handle vouchers when it comes to them
WeWork WebHook, Token, and API Key are all call credentials; do not include prompts, screenshots, or shareable skills. Third-party skills may also read local files or send input to external services. Before installation, verify the source, permission requests, and script content; When deleting, making payments, or sending externally, first verify with small-scale data.
When choosing, you can use a standard sentence: if the main question is "let AI know how to do it," choose Skill; If the main question is "what can AI connect to?", choose MCP; You need to fix the process while accessing external systems, then combine the two.
