Skill file reference.md's contain invalid utf 8 sequence

[sethcarney]

Search keywords

skills

Latest version

• I have tested the latest version #48730

Steps to reproduce

The following skills in the current repo have invalid UTF-8 sequences. This have been flagged by vercels skills.sh as well.
https://www.skills.sh/mui/material-ui/material-ui-theming/security/snyk

Hidden character scan failed for material-ui-styling:

reference.md:1:11 invalid-utf8 U+FFFD invalid UTF-8 byte sequence

Hidden character scan failed for material-ui-tailwind:

reference.md:1:11 invalid-utf8 U+FFFD invalid UTF-8 byte sequence
reference.md:1:25 invalid-utf8 U+FFFD invalid UTF-8 byte sequence
reference.md:16:6 invalid-utf8 U+FFFD invalid UTF-8 byte sequence
reference.md:24:15 invalid-utf8 U+FFFD invalid UTF-8 byte sequence
reference.md:24:43 invalid-utf8 U+FFFD invalid UTF-8 byte sequence
reference.md:24:67 invalid-utf8 U+FFFD invalid UTF-8 byte sequence
reference.md:38:31 invalid-utf8 U+FFFD invalid UTF-8 byte sequence

Hidden character scan failed for material-ui-theming:

reference.md:1:11 invalid-utf8 U+FFFD invalid UTF-8 byte sequence

Current behavior

No response

Expected behavior

These character sequences should be fixed so that the provided skills are not flagged as a prompt injection risk.

Context

No response

Your environment

No response

[siriwatknp]

I think this is on skills.sh to fix, not our side. We are using non-breaking spaces across the repo for brands.

[sethcarney]

I think this is on skills.sh to fix, not our side. We are using non-breaking spaces across the repo for brands.

Thanks for triaging. Quick clarification on scope: this isn't really an external dependency issue. The U+FFFD bytes are committed directly in MUI's own skills/**/reference.md files, so the fix lives entirely in this repo. skills.sh and Snyk are just surfacing them.

Root cause is almost certainly an encoding round-trip: characters like smart quotes or arrows got saved in a non-UTF-8 encoding (likely cp1252) and later replaced with U+FFFD. Locate them with:

​bash grep -rnP '\xEF\xBF\xBD' skills/ ​

Fix is to swap each U+FFFD back to the intended character and re-save as clean UTF-8. Happy to open a PR if useful.