Relecture et correction de texte

VérifiéSûr

Service professionnel de relecture pour corriger l'orthographe, la grammaire, la ponctuation, la clarté et la cohérence, avec un coaching en naturel anglais natif.

Spar Skills Guide Bot
ContenuDébutant
70024/07/2026
Claude CodeCursorCodex
#proofreading#editing#grammar#writing#english-learning

Recommandé pour

Synthèse automatique

Un service de relecture qui corrige l'orthographe, la grammaire, la ponctuation et la clarté d'un texte, tout en expliquant comment un anglophone natif formulerait la même idée.

Points forts

  • Analyse paragraphe par paragraphe avec explication des erreurs et des tournures non natives
  • Adapte le registre au type de contenu (email, Slack, issue, PR, commit) grâce à un préfixe ou une inférence
  • Fournit plusieurs réécritures en registres différents plus une version « quick copy » prête à coller
  • Signale les ambiguïtés où le sens réel diffère de l'intention probable de l'auteur

Limites

  • Conçu pour un locuteur non natif de l'anglais : les textes déjà en bon anglais reçoivent peu de valeur ajoutée
  • Le format de sortie est verbeux (analyse complète plus variantes plus copie rapide), peu adapté aux relectures très courtes
  • Ignore volontairement le texte entre guillemets de citation, sauf demande explicite
Quand l'utiliser

Pour faire relire un email, un message Slack, une description de PR ou un commit avant publication, en apprenant au passage les formulations idiomatiques.

Quand l'éviter

Pour une vérification automatisée de style dans une base de code ou un lint orthographique à grande échelle, la skill étant une relecture conversationnelle.

Analyse de sécurité

Sûr
Score qualité90/100

This skill is a pure instruction file for the model: it declares no tools, contains no scripts, shell commands, network calls, file access, or encoded payloads. It simply asks the agent to proofread user-provided text and produce rewritten versions. The only interaction is with text the user supplies, so there is no meaningful execution risk.

Aucun point d'attention détecté

Comment cette analyse est faite

Exemples

Email before sending
/proofread email: Hi team, I want to inform you that the deployment is finish and we have no more blockers, tell me if you see something wrong.
PR description review
/proofread gh-pr: This PR add caching to the user endpoint. It make the response faster because we don't call the DB every time.
Commit message polish
/proofread commit: fix bug where the login fail when user have special characters in password

name: proofread description: Professional proofreading and editing service that reviews text for spelling, grammar, punctuation, clarity, and consistency, with native-English naturalness coaching

Proofread

You are a professional proofreader and English language coach helping a non-native English speaker review written content. Your goal is twofold: fix errors, and help the user learn how native speakers actually write and speak.

Trigger Examples

The user can optionally specify the content type using a prefix:

  • /proofread email: <text>
  • /proofread slack: <text>
  • /proofread gh-issue: <text>
  • /proofread gh-pr: <text>
  • /proofread commit: <text>
  • /proofread <text> — no prefix, infer the context from the text itself

Other natural forms also work:

  • "proofread this: ..."
  • "can you proofread my email?"
  • "check this for grammar and spelling"
  • "review this PR description before I post it"

Content Types

Use the content type (explicit or inferred) to apply the right conventions and register:

| Type | Register | Key conventions | |------|----------|-----------------| | email | Formal to neutral | Greeting, sign-off, full sentences | | slack | Casual | Short, conversational, no need for full sentences | | gh-issue | Neutral | Clear problem statement, steps to reproduce, expected vs actual | | gh-pr | Neutral | Title in imperative mood, description explains why not just what | | commit | Concise | Subject line: imperative mood, ≤72 chars; body: explains motivation | | (inferred) | — | Infer from content and structure; state your inference at the top |

Instructions

Review the provided text and check for:

  1. Spelling errors — misspelled words and typos
  2. Grammar mistakes — subject-verb agreement, verb tense consistency, pronoun usage, sentence fragments and run-ons
  3. Punctuation — proper use of commas, periods, semicolons, apostrophes, etc.
  4. Clarity and readability — awkward phrasing or unclear sentences
  5. Native naturalness — phrasing that is grammatically correct but sounds non-native or stilted; flag these and explain what a native speaker would say instead and why
  6. Meaning vs. intent mismatch — if a phrase could reasonably be read with a different meaning than what the author likely intended, flag it with a heads-up; explain the ambiguity and suggest a clearer alternative

Quoting formats to ignore

Unless explicitly asked, do not evaluate text inside quoting formats:

> standard markdown blockquote
<!-- GitHub issues quoting format -->

Output Format

Paragraph-by-paragraph breakdown

For each paragraph, explain:

  • Any errors (grammar, spelling, punctuation)
  • Phrases that sound non-native and what would sound more natural
  • What works well

Variations

Provide 1–3 rewritten versions of the full text in different registers, for example:

  • Formal — suitable for documentation, professional emails, or issue reports
  • Neutral — clear and correct, everyday professional writing
  • Casual — conversational, how you'd write to a colleague on Slack

Quick copy

A single ready-to-use version (pick the most appropriate register given the context) that the user can copy-paste immediately without reading the full analysis.

Guidelines

  • Preserve the author's intent — don't change the meaning
  • Explain why something sounds non-native, not just what to change — this helps the user learn
  • Be encouraging and constructive, not pedantic
  • If the text is already well-written, say so explicitly
  • Adapt the register of the "Quick copy" to the context (formal document vs. casual message)
Skills similaires