Skip to content

fix(tools): trim whitespace from skill name to prevent lookup failures#2027

Closed
echoVic wants to merge 5 commits intoQwenLM:mainfrom
echoVic:fix/skill-name-whitespace
Closed

fix(tools): trim whitespace from skill name to prevent lookup failures#2027
echoVic wants to merge 5 commits intoQwenLM:mainfrom
echoVic:fix/skill-name-whitespace

Conversation

@echoVic
Copy link
Copy Markdown
Contributor

@echoVic echoVic commented Mar 1, 2026

Problem

When calling skills with Chinese names (or any name), the model sometimes adds extra whitespace around the skill name, causing the lookup to fail and resulting in an infinite loop.

Fixes #2025

Changes

  • Trim skill name in validateToolParams to handle cases where model adds extra whitespace
  • Update params.skill with trimmed value for consistent usage throughout the tool
  • Add test cases for leading/trailing whitespace and Chinese skill names

Testing

  • Added new test cases covering:
    • Leading space: ' code-review'
    • Trailing space: 'code-review '
    • Both spaces: ' code-review '
    • Whitespace-only rejection: ' '
  • All 23 tests pass

QwenLM#2025)

- Trim skill name in validateToolParams to handle cases where model adds extra whitespace
- Update params.skill with trimmed value for consistent usage throughout the tool
- Add test cases for leading/trailing whitespace and Chinese skill names

Fixes QwenLM#2025
echoVic added 3 commits March 1, 2026 11:43
… API errors (QwenLM#2020)

- Add PDF (application/pdf) to unsupported media types in convertUnsupportedMediaToText
- This prevents 'Invalid value: file' API errors when reading PDF files
- PDF content in tool responses is now converted to explanatory text
- Add test cases for PDF inlineData and fileData conversion

Fixes QwenLM#2020
- Add 'bash' and 'sh' to ToolNamesMigration as aliases for 'run_shell_command'
- Migrate legacy tool names in CoreToolScheduler._schedule() before registry lookup
- Add test case verifying the migration mapping
@tanzhenxin
Copy link
Copy Markdown
Collaborator

@echoVic Thanks for your contribution! This issue is caused by interleaved English/number characters and Chinese Characters, and the model has some trouble with this pattern. Pls checkout issue #1922 for more details.

We prefer not to merge any hotfix patch for this issue, and we are communicating with the related team if this issue can be fixed at the model side once for all.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

调用skill会自动加上空格 导致找不到这个skill无限循环

2 participants