Skill Validation Should Fail on Names Containing Spaces

[hGriff0n]

I'm using langchain, so this might be something that's better for that implementation, especially as this is model specific. Already filed a similar bug there (langchain-ai/langchain-google#1332) but cross-filing it here in case there's a broader issue.

My impression is that gemini is oddly restrictive in how it requires tool definitions to be formatted. One of those requirements is apparently that the tool name not contain spaces. I see all your examples do show a '-' instead of spaces, so possibly my fault.

In the case that this is not model-specific, I would think we could add this to the skill validation on load and prevent an api call that's just going to fail.

[maxvaega]

Hey @hGriff0n ,

Thanks for the Issue, well noted.
while i review the details of the gemini behavior, please note i have just released v0.2.0 that fixes a similar situation with folder names when wrong discovery folders were passed as arguments.
I understand this is not necessarily related to the issue you reported, but was worth mentioning.

Max

[maxvaega]

Hi @hGriff0n ,

the main purpose of this library is to be fully compatible with existing skills, which are supposed to follow rules defined by Anthropic.

In their official documentation, on page Agent Skills, they clearly say:

Field requirements:
name: Must use lowercase letters, numbers, and hyphens only (max 64 characters)

Currently, this is not enforced in skillkit, but needs to be implemented in a future update.
Thanks for pointing this out, looking forward to receive any further feedback from you!

Max