Skip to main content

❓ YingClaw FAQ

👤 Best for: All users

⏱️ Reading time: about 5 minutes

💡 In one sentence: The most common YingClaw questions, answered — skip the small-stuff blockers.

A roundup of the questions and answers that come up most often when using YingClaw.

Install and start​

Q: Installation failed — what now?​

Make sure Node.js is at version 18 or higher, then retry:

node -v          # check version
npm install -g yingclaw   # reinstall

If it still fails, clear the npm cache and try again:

npm cache clean --force
npm install -g yingclaw

Q: Installed on Windows 11 but the client won't open — no window or web page shows up?​

Symptom and cause: Windows 11 shows a "Smart App Control blocked a potentially unsafe app" prompt. Even after you click through, the client still fails to open and no web page appears. The cause is Smart App Control intercepting an unsigned program.

Smart App Control warning - step 1

Steps to fix it:

(1) Open Settings → Privacy & security → Windows Security.

Disable Smart App Control - step 1

(2) Choose App & browser control.

Disable Smart App Control - step 2

(3) Open Smart App Control settings.

Disable Smart App Control - step 3

(4) Switch the option to Off.

Disable Smart App Control - step 4

(5) Once changed, launch the YingClaw client again.

Q: "Model unavailable" after starting?​

Check the network connection and confirm the configured AI model API is reachable. For a local model, make sure the service is up and accessible.

Q: Which operating systems are supported?​

OSMinimum version
Windows10 and above
macOS11 (Big Sur) and above
LinuxMainstream distros (Ubuntu 20.04+, Debian 11+)

Q: Linux client online update keeps failing — what now?​

There is no quick fix available right now. The workaround is to redownload the latest full client package and install it over the existing one. As shown below, open System Update in the YingClaw page and click Confirm install.

Linux client update fix

Using the features​

Q: Where is memory data stored?​

  • Windows: %USERPROFILE%\.YingClaw\workspace\
  • macOS / Linux: ~/.YingClaw/workspace/

Memory lives in MEMORY.md and the memory/ daily notes directory, persisted to disk.

Q: How is memory privacy protected?​

All memory data stays on the local machine and is never auto-uploaded to any external server. You can edit or delete memory files any time.

Q: Skill vs plugin — what's the difference?​

A Skill is YingClaw's reusable procedural-knowledge module: full step-by-step workflow, common pitfalls, and verification. Unlike a code plugin, a skill focuses on teaching the AI an executable workflow.

Q: Can scheduled tasks run in parallel?​

Yes. Each scheduled task runs independently; you can set a timeout and an auto-disable condition.

Q: Does browser automation need a GUI?​

No. YingClaw ships with a headless browser, so it works fine in pure command-line environments (e.g. a remote server).

Agents and orchestration​

Q: What's the difference between a sub-agent and an executor?​

  • Main agent (the YingClaw you chat with): understands the request, plans the task, and dispatches work
  • Sub-agent: handles a specific subtask with its own toolset and context window

When the sub-agent finishes, the result is returned to the main agent for summarisation.

Q: Can a sub-agent see my memory?​

By default, a sub-agent inherits core and user memory. To isolate specific information, restrict the available skills when dispatching via the skills parameter.

Q: How many sub-agents can run in parallel?​

No hard limit; the real ceiling is the model API concurrency and machine resources. We recommend 2-5 in flight at once, following the natural split of the task.

Security and permissions​

Q: Will YingClaw run dangerous commands?​

Built-in safety policy automatically blocks these high-risk operations:

  • Recursive force delete (rm -rf)
  • Disk formatting (format, mkfs)
  • Silent delete of the system drive
  • Registry modifications

Medium-risk operations will require explicit confirmation.

Q: Can I block specific commands?​

Yes. Use the safety policy config to set command blacklists or whitelists, and precisely control what is allowed.

Q: How should I handle sensitive info?​

  • Credentials, API keys and similar secrets should not be stored in memory
  • Tool output containing credentials is automatically redacted before being returned
  • Put API keys in a .env file and let YingClaw manage them securely

Advanced usage​

Q: How do I make YingClaw understand my project better?​

  1. Store project decisions and coding conventions in core memory
  2. Create custom skills to wrap project-specific workflows
  3. Track progress in daily notes

Q: How do I switch between multiple projects?​

YingClaw organizes context by workspace. Different projects can use different workspaces and stay isolated.

Q: Which large models are supported?​

Any mainstream model reachable through a standard API — depends on your API configuration. You can use model routing to assign different models to different scenarios.