fix(gh-body-audit,gh-body-conventions): state where two math rules apply

[ultimatile]

Summary

gh-body-conventions § Math forbade raw Unicode math characters "in prose" and forbade \operatorname with no qualifier, so neither rule said whether it reached text inside a code span, a fenced code block, or math. gh-body-audit step 2 filled that gap with its own list of positions: it dismissed a math-scan hit inside a fenced code block, inside an inline code span, or in prose that only names the construct. Backtick-delimited inline math is written as a code span and a math fence is a fenced code block, so the list dismissed \operatorname in both, which is the text the rule is about. Each rule now names the text it reaches by how GitHub renders it, and step 2 points at the rules in place of the list.

Closes #242

Changes

• § Math forbids raw Unicode math characters in text GitHub renders as neither code nor math.
• § Math forbids \operatorname in text GitHub renders as math.
• gh-body-audit step 2 flags a glyph or \operatorname hit that sits in the text its § Math rule prohibits it in, and marks any other such hit not applicable. It holds no list of positions.
• The header comment and usage text of skills/gh-body-audit/body-math-scan.sh and the header comment of skills/gh-body-audit/test-body-math-scan.sh name the three scanned classes — Unicode math glyphs, \operatorname, and inline math wrapped in a longer code span — without calling every flagged construct forbidden, and the scan's header leaves the judgment of whether a hit sits in text its rule covers to skills/gh-body-audit/SKILL.md. The scan's regex and exit codes are as they were.
• metadata.version in .claude-plugin/marketplace.json is 2026.10.3.

Impact

Four cases get a different outcome from the audit's math-scan step:

• \operatorname in backtick-delimited inline math or in a math fence was dismissed and is flagged.
• A glyph in dollar-delimited inline math was flagged and is not applicable. The position list had no entry for it, while "in prose" already left it outside the rule.
• A glyph in text that is neither code nor math and that only names the glyph was dismissed and is flagged. A code span around the glyph clears it.
• \operatorname in text that is neither code nor math and that does more than name the macro was flagged and is not applicable, since the math renderer never receives it.

file-issue and file-pullreq apply § Math by reference and take the reworded rules with no edit of their own.

Verification

• gh api -X POST /markdown in gfm mode, given one body holding each form, returned a <math-renderer> element for backtick-delimited inline math, dollar-delimited inline math, and a math fence; a <code> or <pre> element for a code span, a tex fence, and backtick-delimited inline math wrapped in a longer code span; and neither for paragraph text and a table header cell. The three classes the rules name are the ones GitHub's renderer produces.
• A fresh-context agent, given skills/gh-body-audit/SKILL.md and skills/gh-body-conventions/SKILL.md as this PR leaves them, ran step 2 on a body that places glyphs and \operatorname in a heading, paragraph text, code spans, a text fence, both inline-math forms, and a math fence, with one line also holding inline math wrapped in a longer code span. The scan printed 14 lines, which the agent split into 17 hits: it flagged the three glyphs in text that is neither code nor math and the three \operatorname in math, and marked the other 11 hits not applicable. That covers the four changed cases above.
• bash skills/gh-body-audit/test-body-math-scan.sh printed All checks passed.
• The pre-commit mdformat hook reported Passed on the edited files.

Notes

• The math scan does not report $$ display math, and step 2 has no rule for a $$ that is not display math; body-math-scan: report $$ display math #243 tracks both.

[Copilot]

Copilot review overview

🟢 Approval recommended

The updated guidance consistently resolves the ambiguity from issue #242 without altering scanner behavior.

Review effort: Balanced
Findings: None

What changed in this PR

Clarifies where GitHub body math rules apply and aligns audit guidance with the conventions SSOT.

Changes:

• Scopes Unicode glyph restrictions to non-code, non-math text.
• Scopes \operatorname restrictions to rendered math.
• Updates scanner documentation and bumps the plugin version.

File	Description
.claude-plugin/​marketplace.json	Bumps version to 2026.10.3.
skills/​gh-body-conventions/​SKILL.md	Defines each math rule’s rendering scope.
skills/​gh-body-audit/​SKILL.md	Delegates hit applicability to the corresponding convention.
skills/​gh-body-audit/​body-math-scan.sh	Clarifies scanner scope without changing behavior.
skills/​gh-body-audit/​test-body-math-scan.sh	Corrects the test harness contract description.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.