fix: preserve Skills upload file paths

[Genmin]

Fixes #1807.

This keeps the existing safe default for multipart uploads: filenames inferred from File, Response, stream path, or filename are still stripped down to their basename unless a caller opts out.

The Skills API is the exception because its upload contract requires files to live under a single top-level directory, such as my-skill/SKILL.md. skills.create() and skills.versions.create() now opt into preserving filename paths when constructing multipart form data.

Validation:

• yarn test tests/form.test.ts

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a20e148b58

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Limit path-preserving mode to explicit file names

When stripFilename === false, this returns the raw value for every source (name, url, filename, and path), but Skills now enables that mode by default in both create calls. As a result, Response uploads use full URLs and stream uploads can use full filesystem paths (e.g., from fs.createReadStream(path.resolve(...))), which can violate the Skills “single top-level directory” filename contract and leak local paths in multipart metadata. Path-preserving mode should be restricted to caller-provided logical names, while inferred url/path values should still be reduced to a basename.

Useful? React with 👍 / 👎.