Skip to main content

❓ FAQ

👤 Who it's for: All users
⏱️ Read time: ~5 minutes
💡 In one line: High-frequency questions in one place — don't get stuck on small things.

Common questions and answers about using YingClaw.

Installation & Startup​

Q: Installation failed. What should I do?​

Make sure Node.js >= 18, then retry:

node -v              # Check version
npm install -g yingclaw   # Reinstall

If it still fails, clear npm cache and reinstall:

npm cache clean --force
npm install -g yingclaw

Q: It says "model unavailable" after starting?​

Check network connectivity to your configured AI model API. For local models, verify the service is running and reachable.

Q: Which operating systems are supported?​

OSVersion Requirement
Windows10 and above
macOS11 (Big Sur) and above
LinuxMajor distributions (Ubuntu 20.04+, Debian 11+)

Features & Usage​

Q: Where is memory data stored?​

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

Memory is persisted as MEMORY.md and daily notes under the memory/ directory.

Q: How is memory privacy protected?​

All memory data is stored locally and never automatically uploaded to external servers. You can manually edit or delete memory files anytime.

Q: What's the difference between skills and plugins?​

Skills are reusable procedural knowledge modules containing complete operational steps, common pitfalls, and verification methods. Unlike code plugins, skills focus on teaching AI executable workflows.

Q: Can scheduled tasks run concurrently?​

Yes. Scheduled tasks run independently and support timeout settings and auto-disable conditions.

Q: Does browser automation require a GUI?​

No. YingClaw has a built-in headless browser that works in pure CLI environments (such as remote servers).

Agents & Orchestration​

Q: What's the difference between sub-agents and executors?​

  • Main Agent (the YingClaw you talk to): Understands requirements, plans tasks, dispatches work
  • Sub-Agent: Executes specific subtasks with its own toolset and context window

Results automatically flow back to the main agent when a sub-agent completes.

Q: Can sub-agents see my memory?​

By default, sub-agents inherit core and user memory. To isolate specific information, limit available skills via the skills parameter when dispatching.

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

No hard limit, but actual throughput depends on model API concurrency and machine resources. Split tasks by natural boundaries — 2-5 parallel agents is typically the sweet spot.

Security & Permissions​

Q: Will YingClaw execute dangerous commands?​

The built-in security policy automatically blocks the following high-risk operations:

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

Medium-risk operations require explicit confirmation.

Q: Can I restrict which commands can be executed?​

Yes. Configure allowlists/blocklists in the security policy to precisely control executable operations.

Q: How is sensitive information handled?​

  • Credentials, keys, and other secrets should NOT be stored in memory
  • Tool outputs containing credentials are automatically redacted before returning to the user
  • Store API keys in .env files, managed securely by YingClaw

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 for project-specific workflows
  3. Track progress through daily notes

Q: How to switch between multiple projects?​

YingClaw organizes context by workspace. Use different workspaces for different projects — they remain fully independent.

Q: Which LLMs are supported?​

Any major LLM accessible via standard API endpoints, depending on your API configuration. You can route different scenarios to different models through model routing config.

Running into issues?​