Skill description limit may be checking byte length instead of character count

Resolved 💬 3 comments Opened Dec 8, 2025 by Mochi-frontend Closed Dec 12, 2025
💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

What version of Codex is running?

0.66.0-alpha.10

What subscription do you have?

Business

Which model were you using?

gpt-5.1-codex-max (reasoning xhigh, summaries auto)

What platform is your computer?

Darwin 24.5.0 arm64 arm

What issue are you seeing?

When loading a skill with a Japanese description, I get this error:

invalid description: exceeds maximum length of 500 characters

However, my description is clearly under 500 characters.

What steps can reproduce the bug?

  1. Create a SKILL.md file with a Japanese description that is under 500 characters but over 500 bytes
  2. Run codex in the directory containing the skill
  3. The skill fails to load with the error above

What is the expected behavior?

A description under 500 characters should pass the limit documented in docs/skills.md:
https://github.com/openai/codex/blob/main/docs/skills.md

Additional information

Possible cause: Rust's str::len() returns byte length, not character count. If the validation uses .len(), Japanese/Chinese/Korean descriptions might hit the limit much earlier than expected.

Note: I'm a beginner in engineering, so I may be mistaken. I apologize if this is not actually a bug.

View original on GitHub ↗

3 Comments

etraut-openai contributor · 8 months ago

Thanks for the bug report.

The skills feature is still under active development. When it's ready for use, we'll make an announcement and/or add it to the release notes.

etraut-openai contributor · 8 months ago

This will be fixed in the next release.

Mochi-frontend · 8 months ago

@etraut-openai
Thanks for the quick response! That helps a lot :)