Accept Letters in Recorded Names in the Charset Rule and Prose Gate #2100

The charset rule treats every non-ASCII character outside its three tiers as a finding, and prose_lint.py reports an accented letter as charset-unknown ("is in no tier - classify it in GOVERNANCE.md"). That is right for typography, and wrong for the spelling of a person's or place's name that a document records as data.

Maintainer ruling: the actual spelling of a name being recorded is not subject to the ASCII default. Folding it to ASCII records a different name.

Constructed reproduction: a Markdown line such as - b2 Zoë: married 1702 René Müller reports three charset-unknown findings (U+00EB, U+00E9, U+00FC). There's no per-line escape. The only repo-side knob is .github/prose-gate-excludes, which drops the whole file from every prose rule.

Current workaround downstream: write the letter as an HTML entity (René), which renders exactly and passes the gate. It keeps the raw Markdown hard to read, so treat it as a stopgap.

Asked for:

• The charset rule in GOVERNANCE.md "Documentation Style Conventions" (and the comment-and-doc-style skill that carries it) states that letters in a recorded name keep their spelling.
• The gate accepts letters (Unicode category L, Latin script at least) instead of reporting them as charset-unknown, while tier 1 typography stays a finding. Precomposed letters are the realistic case, and a combining mark on its own would still be worth reporting.

🤖 Generated with Claude Code