Skip to content
Change language: EN
My learningAdd to Chrome
Track progress 0%

Lesson 4 5 min

Give it skills

Skills are what an agent can do. Turn on only what its job needs, with a model that can call tools.

The six skills

Turn skills on in the agent builder’s Skills tab.

Skill What the agent gets
Browser control Its own tab group to read, click and type in. Needs the extension.
Web search Searches the web in a background tab and reads the results. Needs the extension.
Workflows Each chosen workflow becomes a tool. The agent passes it some text and gets back what its last steps produced.
On-device AI Image, text and audio models that run on your device.
Teammates Can ask the teammates you pick to do part of the work, each with its own model, skills and rules.
MCP The tools of each MCP server you add. For a server that needs a login, pick a saved credential; only a reference to it is stored on the agent. Each MCP call follows the MCP rule, with a row per server on the Rules tab.
The Skills tab of the agent builder for Price checker: switches for Browser control, Web search, Workflows, On-device AI (on, Allowed), Teammates and MCP servers. In the web app, Browser control and Web search say they need the AWFlow extension.The Skills tab of the agent builder for Price checker: switches for Browser control, Web search, Workflows, On-device AI (on, Allowed), Teammates and MCP servers. In the web app, Browser control and Web search say they need the AWFlow extension.
Turn a skill on

A workflow that pauses for an approval or a wait can’t finish inside the call, so the agent is told it failed. Teammates don’t see the chat, so the agent gives them a complete task.

The model matters

Skills need a model that can call tools. Chrome’s built-in Gemini Nano handles a few simple skills (up to 8 tools at a time, no browser control); with a chat-only model, such as some Ollama models, the agent’s skills are off. When a skill can’t be used (no extension, a deleted workflow, an MCP server that doesn’t answer), a banner above the composer says so, with a fix.

Limits that protect you

  • What a tool brings back from outside is given to the model as data to read, never as instructions.
  • Each tool call has a time limit: 60 seconds, or 2 minutes for a workflow. The clock pauses while the agent waits for your approval.
  • Very long results are trimmed, and exactly the same call repeated is skipped.

Key takeaways

  1. Browser control and Web search need the AWFlow extension.
  2. Each workflow you choose becomes a tool the agent can call with some text.
  3. Skills need a tool-calling model; Gemini Nano handles a few simple ones.