Fix grammar and spelling errors in: $ARGUMENTS
If no files are specified, propose the most recent prose target from session context (recent read/edit) and confirm. Ask only when none is detectable.
Rules
- Fix only spelling and grammar errors
- Do not rephrase, reword, or improve style
- Do not translate anything — keep the language of each sentence as-is
- Do not modify code blocks, MDX tags, frontmatter, or custom syntax
- Preserve formatting, structure, meaning, technical terms, and anglicisms
- Handle multilingual content naturally (e.g., French text with English technical terms)
Workflow
Single file — Read, apply corrections with Edit using minimal diffs, report the count.
Multiple files — Spawn one subagent per file in a single message so they run in parallel. Each subagent gets the same rules above and returns its file's correction count. Collect results when all complete and produce the combined report.
Grammar checks are per-file independent work with no shared state, so parallelism is pure latency win — N files finish in ~1× file time instead of N×.
Examples
- English typos — "Thier cat wos hungrey" → "Their cat was hungry" (3 typos fixed)
- Multilingual + accents — "J'ai deployé la fonction useEffect" → "J'ai déployé la fonction useEffect" (accent restored; code identifier untouched)
- Code blocks stay verbatim — any fenced block is preserved even if spelling looks odd inside
Output
One line per file, aggregated at the end for multi-file runs:
fix-grammar: {filename} — {n} corrections