[FanWu-ai]

a1_to_rowcol currently accepts only a prefix of its input. For example, B2!D4 becomes (2, 2), and worksheet.update_acell('B2!D4', value) can update B2 rather than rejecting an unsupported sheet-qualified label. Ranges, appended text, decimal-looking suffixes and trailing newlines are silently truncated too.

Require a full regex match for this single-cell conversion and raise the existing IncorrectCellLabel before any read/write dispatch. Valid labels and the separate range parser retain their behavior. Regressions cover seven malformed labels and the public Cell/acell/update_acell paths.

Validation:

• Pristine full suite: 166 passed; final: 168 passed on every Python 3.8–3.14 interpreter, no skips.
• Eight original regression failures; 27,000 valid conversions unchanged.
• Required lint, type/security checks, current pyupgrade, strict Sphinx docs, wheel/sdist and strict metadata validation pass.
• The pinned Black rules were run serially on all 18 files discovered by Black itself because this environment blocks its multiprocessing socket; no rules or discovery filters changed.
• Current pyupgrade requires the included small identity-comprehension normalization in Worksheet.findall; no dependency/checker pins changed.

The advisory pip-audit step (continue-on-error in upstream CI) still reports pre-existing vulnerabilities in pinned dependencies; it is not claimed green. Tests use the existing offline recordings without new credentials or live API calls. Other operating systems and hosted CI have not run locally.

AI assistance: OpenAI Codex was used to investigate, implement and test this fix. No human review is claimed.