Skills execution exception

[zonglinZhang]

Pre-checks

• I have searched existing issues and this is not a duplicate.

Deployment Method

Docker

Steps to Reproduce

When the Agent is executing skills, it seems to default to executing in the workspace directory, and the skills directory is not found.

Expected vs Actual Behavior

Install a skill, then have the agent execute it

Logs / Screenshots

No response

[Clawiee]

Hi @zonglinZhang,

Thanks for reporting this issue! I can see the problem - when executing skills, the agent is looking for the skills directory relative to the workspace root, but the path resolution might be incorrect.

This appears to be a path resolution issue in the skill execution logic. The skills should be located at skills/ directory within the agent's workspace. Let me investigate the source code to understand the exact cause.

Could you provide more details about:

1. Which skill were you trying to execute?
2. How did you install the skill (via install_skill tool or manually)?
3. What's your deployment configuration?

I'll look into this and provide a fix soon.

Best regards,
Clawiee (Clawith Product Assistant)

[zonglinZhang]

Hi @zonglinZhang,  你好，

Thanks for reporting this issue! I can see the problem - when executing skills, the agent is looking for the skills directory relative to the workspace root, but the path resolution might be incorrect.感谢您报告此问题！我可以看到问题 - 在执行技能时，代理会搜索与工作区根目录相对应的技能目录，但路径解析可能不正确。

This appears to be a path resolution issue in the skill execution logic. The skills should be located at skills/ directory within the agent's workspace. Let me investigate the source code to understand the exact cause.这似乎是一个技能执行逻辑中的路径解析问题。技能应该位于代理工作区内的 skills/ 目录中。让我检查源代码，以了解确切的原因。

Could you provide more details about:能否提供更多关于：

1. Which skill were you trying to execute?你试图执行哪项技能？
2. How did you install the skill (via install_skill tool or manually)?您是如何安装技能的（通过 installSkill 工具或手动安装）？
3. What's your deployment configuration?您的部署配置是什么？

I'll look into this and provide a fix soon.我会研究一下，很快就会提供修复。

Best regards,  致上 Clawiee (Clawith Product Assistant)Clawiee（Clawith 产品助理）

My skills were developed by myself. In the company's management interface, I uploaded the skill code by uploading files.
and I used docker deploy Clawith

[Clawiee]

Hi @zonglinZhang, thank you for reporting this bug! 🔍

This appears to be a skill execution path issue where the agent is looking for skills in the workspace directory instead of the skills directory. This could be related to how the skill runtime resolves file paths.

Initial analysis:

• The error suggests the skill execution context is not properly set
• The agent might be using the wrong working directory when executing skills
• This could be in the skill loader or execution engine

We'll investigate this further. Could you please provide:

1. Which skill you were trying to execute?
2. The full error message from the logs?
3. Your deployment method (Docker/Source)?

Thank you for helping us improve Clawith!

---

你好 @zonglinZhang，感谢报告这个 bug！🔍

这看起来是一个技能执行路径问题，agent 在 workspace 目录而不是 skills 目录中查找技能。这可能与技能运行时如何解析文件路径有关。

初步分析：

• 错误表明技能执行上下文未正确设置
• agent 执行技能时可能使用了错误的工作目录
• 问题可能在技能加载器或执行引擎中

我们会进一步调查。能否请你提供：

1. 你尝试执行的是哪个技能？
2. 日志中的完整错误信息？
3. 你的部署方式（Docker/源码）？

感谢你帮助我们改进 Clawith！

[wisdomqin]

Hi @zonglinZhang, thanks for reporting this!

We've identified and fixed the root cause. The execute_code tool was setting its working directory to {agent_root}/workspace/, which is a subdirectory of the agent's root. This meant that when code tried to reference files in skills/ (e.g., open("skills/my-skill.md")), it would look for workspace/skills/... which doesn't exist.

Fix (commit ca07269): We changed the execute_code working directory from {agent_root}/workspace/ to {agent_root}/ (the agent root). Now code execution can directly access:

• skills/ — your skill definitions
• workspace/ — working files and uploads
• memory/ — agent memory
• soul.md — agent personality

The fix has been deployed to our production environment. If you're self-hosting with Docker, you can update by pulling the latest code:

git pull origin main
docker cp backend/app clawith-backend-1:/app/
docker exec clawith-backend-1 find /app -name "__pycache__" -exec rm -rf {} + 2>/dev/null
docker compose restart backend

Please let us know if this resolves the issue for you!