Description length counts UTF-8 bytes but reports "characters" — multibyte descriptions rejected below 1024 chars

[BoneLiu]

Summary

The 1024-character description limit is enforced with Go's len(desc), which counts UTF-8 bytes, while the error message and the spec both speak of characters. Multibyte (CJK, emoji) descriptions are therefore rejected well below 1024 real characters.

Evidence

structure/frontmatter.go:47:

results = append(results, ctx.Errorf("description exceeds 1024 characters (%d)", len(desc)))

Real-world case from our skill tree:

• a SKILL.md description measuring 984 characters (compliant) and 1200 UTF-8 bytes (rejected)
• the skill is flagged description exceeds 1024 characters (1200)

The spec's reference implementation enforces the same limit with Python len(), i.e. characters (skill-creator/scripts/quick_validate.py:83).

The existing test cannot catch this because its long-description fixture is ASCII (structure/frontmatter_test.go:107, strings.Repeat("x", 1025)) — bytes == chars there.

Suggested fix

results = append(results, ctx.Errorf("description exceeds 1024 characters (%d)", utf8.RuneCountInString(desc)))

plus a regression test with a CJK description, e.g. 342 CJK chars = 1026 bytes = 342 characters (must pass).

Happy to send a PR if you tell me which branch to base on.

[dacharyc]

Thanks for filing this! I've taken a small poll of open source harnesses I can inspect, and it looks like the "character" interpretation is ambiguous and implemented differently across different harnesses. I think your interpretation is the most correct one, though, so I've opened a PR with a fix and will merge and release shortly.