Cate CLI Skill
Drive browsers, terminals, editors, and panels from the command line inside Cate's infinite canvas.
The skill requires explicit CLI permission in settings, indicating a permission model. Documentation states that browser and terminal commands reject panels of wrong type, preventing accidental control. Terminal input requires verification of panel ID and screen. However, there is no explicit user confirmation mechanism for dangerous operations, and permission granularity may be insufficient. No red lines found, but static review cannot verify actual behavior, so full marks not awarded.
The document provides command examples and execution steps, but lacks automated tests or error handling details. Failure modes exist (e.g., stale refs, panel type mismatches), but error feedback mechanisms are not described. Static reading cannot verify reproducibility, so high score not given.
The skill's goal is clear (drive Cate's browser, terminal, editor panels) and scenarios are clear. Documentation declares unsupported features (e.g., native session/CDP switching, upload/download paths), helping define boundaries. But it lacks explicit non-fit scenarios and does not mention Chinese language support or mainland China network reachability, deducting points.
Documentation is well-structured with sections, examples, and steps. License and version info are clear in package.json. But there is no dedicated troubleshooting guide, FAQ, or changelog. The skill itself lacks versioning and update path info, making governance incomplete.
Documentation claims to complete core tasks like browser automation and terminal operations, with example outputs (e.g., screenshot paths). But static reading cannot verify if outputs are directly usable, and no comparison with alternatives is provided. Given static assessment, baseline score given.
The repository contains CI workflows and test config files, but evidence that the skill itself is tested is insufficient. Example commands may work, but no third-party execution evidence or independent test results are attached. Static review limits verifiability.
- The skill may depend on overseas services (e.g., GitHub Actions, Sentry) that could be restricted in mainland China. Ensure network reachability.
- The permission model might not be fine-grained; users should review Cate's CLI permission settings to ensure least privilege.
What does this skill do, and when should you use it?
This skill enables AI agents to control a Cate workspace via its CLI. It allows manipulation of native panels, including browser automation (using agent-browser syntax), terminal typing/reading, editor file opening, and panel management. The skill requires the Cate desktop app itself and the corresponding CLI permission setting. It isolates selected panels per terminal session and provides an inspect-act-wait browser workflow.
The skill executes these operations: lists and selects panels on the canvas; automates browser pages using native agent-browser commands (like open, snapshot, fill, click, wait); manages browser tabs and viewports; reads terminal output and types into terminals (without auto-Enter); opens files at specific line numbers in the editor; creates and closes panels. All operations are driven by Cate's built-in CLI.
- A developer debugging a web app in Cate uses browser snapshots to extract accessibility tree and click elements.
- A QA engineer runs end-to-end tests in the canvas, waiting for URL changes and capturing screenshots.
- A developer uses a terminal panel to run commands, reading output first before sending input to avoid mishaps.
- A project manager quickly opens specific files and lines for review within a large project.
- In multi-agent collaboration, panel selection isolates targets to prevent interference between agents.
What are this skill's strengths and limitations?
- Deep integration with Cate's canvas environment, providing unified browser and terminal control.
- Browser automation leverages mature agent-browser syntax, supporting rich operations.
- Panel selection is session-isolated, ideal for multi-agent parallel work.
- Open-source MIT license, freely usable and modifiable.
- Only works within Cate; not usable in other IDEs or standalone.
- Some native browser features (like CDP switching, upload/download paths) are unavailable and require Cate lifecycle commands.
- Requires manual CLI permission setup, otherwise commands may be rejected.
- Maintenance status unclear; may lack long-term support.
How do you install this skill?
Download and install the Cate desktop app from GitHub Releases (macOS, Windows, Linux). This skill is bundled with the app; no extra installation is needed.
How do you use this skill?
Ensure the CLI permission is enabled in Cate settings. Run cate commands in a Cate terminal or agent shell, e.g., first cate panel list to view panels, then cate panel set <id> to select a target, and use commands like cate browser snapshot -i to automate the browser.