Fix s07 skill frontmatter parsing edge cases

[Bill-Billion]

Summary

• Tighten s07 SKILL.md frontmatter parsing to require line-delimited --- fences.
• Ignore non-dict YAML metadata and fall back when name or description is empty.
• Sync the s07 code snippet in Chinese, English, and Japanese READMEs.

Fixes #430

Verification

• python3 -m py_compile s07_skill_loading/code.py
• Boundary-case parser smoke test with a fake yaml.safe_load shim, covering normal metadata, non-frontmatter text, scalar YAML, missing description, and empty name.
• git diff --check
• Temporary Python 3.12 venv: pytest -q -> 24 passed, 30 subtests passed

[vercel]

@Bill-Billion is attempting to deploy a commit to the crazyboym's projects Team on Vercel.

A member of the Team first needs to authorize it.

[dethan3]

The fix in s07_skill_loading/code.py looks correct, but this tutorial is structured as an incremental codebase. Since s07 introduces the skill loading mechanism, later chapters that build on it should be checked and kept in sync.

I noticed that s08_context_compact/code.py still has the older frontmatter parsing logic, and s20_comprehensive/code.py may need the same update as well.

Before merging, please check s08 through s20 and synchronize any skill loading / frontmatter parsing code that still builds on the s07 implementation. It would also be helpful to mention which later chapters were checked.