Présentation des prompts et skills

Prompts IA sélectionnés pour des usages professionnels, créatifs et techniques.

Works well with any strong creative-reasoning model (Claude, GPT-5-class, Gemini 2.5 Pro). Needs a model willing to generate genuinely restrictive rules rather than safe, decorative ones — weaker models tend to invent constraints that don't actually rule anything out.creativeYour team ran two open brainstorming sessions for a cold brew coffee brand's launch campaign, generated 40 concepts, and a blind test showed people couldn't distinguish any of them from three competitor brands already on the shelf.

Le Générateur de Contraintes Créatives : Transformez un Projet Bloqué en Règles qui Forcent de Nouvelles Idées

Un prompt structuré qui diagnostique pourquoi un brainstorming a convergé vers des idées génériques, puis invente cinq règles délibérément restrictives conçues pour briser ce défaut précis — transformant le blocage créatif en un ensemble classé de concepts à prototyper.

10/20/2026

ai-promptscreative-strategy

Works best with strong-reasoning models (Claude Opus, Claude Sonnet, GPT-5-class, Gemini 2.5 Pro). Weaker models tend to list generic caveats instead of scenario-specific alternative explanations.Analyse de DonnéesYour marketing team just told leadership that a new onboarding email sequence increased 30-day retention by 12%, based on a before/after comparison, and the company is about to reallocate a team's entire budget on the strength of that one chart.

Le Testeur de Résistance Corrélation vs Causalité : Mettez une Affirmation Basée sur des Données à l’Épreuve Avant d’Agir

Un prompt structuré qui force une IA à lister toutes les explications alternatives à un motif de données, à classer les preuves qui les distinguent, et à recommander le diagnostic le moins coûteux — avant d’engager un budget sur une histoire causale potentiellement fausse.

10/20/2026

ai-promptsdecision-making

Works best with Claude Sonnet 5.5 or GPT-5.4 — the three-step diagnose/reconcile/replan structure needs strong instruction-following to stay sequential rather than collapsing into a single pass. Gemini 3 Pro also handles it well; smaller/faster models tend to skip the diagnosis step and jump straight to replanning.productivityYou are a product manager (or any individual contributor juggling multiple workstreams) who ends most Fridays with 2-3 planned priorities still undone, and every Monday you face the same choice: silently carry everything forward, or sit down and actually figure out what happened before deciding what is realistic this week.

Le Réconciliateur de Dérapage Hebdomadaire : Transformez le Plan de la Semaine Dernière vs la Réalité en Priorités Réalistes pour Cette Semaine

Un prompt diagnostiquer-réconcilier-replanifier qui découvre pourquoi le plan de la semaine dernière a dérapé avant de vous laisser construire celui de cette semaine — pour que vous arrêtiez de recréer la même liste irréaliste.

10/19/2026

Works best with Claude Sonnet 5.5 or GPT-5.4 — both have strong grounding in venture financing conventions and can reliably distinguish market-standard terms from genuine outliers. Treat the output as preparation for a conversation with actual counsel, not a substitute for one — no model should be the only check on a legal document before you sign it.Finance et NégociationYou're a first-time founder who just received a term sheet from a lead investor for your seed round, with a Tuesday deadline to respond, and you've never negotiated a priced round before — your lawyer is available but billing by the hour, and you want to walk into that call already knowing where the real fights are.

Le scanner de signaux d'alerte de term sheet : repérez les clauses défavorables au fondateur avant de signer

Un prompt d'IA structuré qui transforme un term sheet confus en une grille de notation clause par clause et une liste hiérarchisée de ce qui mérite vraiment d'être négocié.

10/18/2026

venture-capitalai-prompts

Works best with Claude Sonnet 5.5 or GPT-5.4 — both handle long style guides and nuanced tone matching reliably. Works adequately with Gemini 3 Pro, but it tends to over-flag minor stylistic variance as violations; tighten the Constraints section's issue cap if using Gemini.marketingYour startup just brought on three freelance copywriters for a product launch sprint, and marketing ops has fifteen drafts to clear before anything ships — but the only brand reference anyone has is a scattered, two-year-old Google Doc that none of the freelancers have fully read.

L'auditeur de cohérence de voix de marque : évaluez tout brouillon face à votre guide de style

Un prompt d'IA structuré qui transforme un jugement vague « est-ce que ça nous ressemble » en une grille de notation, une liste de corrections hiérarchisées et un verdict publier/attendre.

10/18/2026

ai-promptscopywriting

Works best with Claude Sonnet 5.5 or GPT-5.4, both of which handle the prioritization logic (picking the highest-leverage combinations rather than covering every item) more reliably than models that default to exhaustive coverage. Gemini 3 Pro also works but may need an explicit reminder to limit the action plan to 7 items rather than listing one action per SWOT bullet.Business et stratégieYour team just finished a quarterly strategy offsite and produced a SWOT analysis with 20+ bullet points across four quadrants, but the document is about to sit in a shared drive untouched the moment everyone goes back to their day jobs.

Le Convertisseur SWOT-en-Action : Transformez une Analyse Stratégique en Plan de 30 Jours avec Responsables et Échéances

Un prompt structuré qui transforme une analyse SWOT en un petit plan d’action attribué et mesurable sur 30 jours.

10/17/2026

Works best with Claude Sonnet 5.5 or GPT-5.4, both strong at mining specific, non-generic detail out of unstructured resume text. Gemini 3 Pro also works, but may need an explicit reminder to include the closing Story Inventory section, which it sometimes drops when the competency list runs long.CarrièreYou have a final-round interview in three days for a senior role. You have 12+ years of relevant experience, but when you try to recall specific stories under time pressure, you blank, or you default to the same two anecdotes for every question regardless of what's actually being asked.

Le Sprint d’Entretien Comportemental : Transformez une Offre d’Emploi en Histoires au Format STAR

Un prompt structuré qui transforme une offre d’emploi et votre CV en un ensemble pré-cartographié d’histoires d’entretien au format STAR.

10/17/2026

Works best with Claude Opus 4.5 or GPT-5.1, both of which can hold an entire long report in context at once rather than summarizing it in chunks — chunked summarization is what causes cross-section contradictions and caveats to get lost. Avoid routing this through a model with a short context window; truncating the source report defeats the entire point of a faithful compression.Rédaction et CommunicationA VP of Operations receives a 24-page vendor due-diligence report late on a Friday, with a budget committee meeting scheduled for Monday morning at 8am where the CFO needs to approve or reject a vendor contract renewal, and there is no time for the committee to read the full report before deciding.

Le Compresseur de Note Exécutive : Transformez un Rapport de 20 Pages en un Mémo d'Action d'Une Page

Un prompt IA structuré qui compresse un long rapport en une note d'une page sur laquelle un décideur peut agir — en préservant les réserves et contradictions de l'original plutôt qu'en les lissant en une fausse confiance.

10/16/2026

productivitydecision-making

Works well with Claude Opus 4.5 or GPT-5.1 for the structured cross-referencing this requires — both hold attribution steady across long combined source text without losing track of which claim came from which source. Gemini 3 Pro works fine for shorter source sets (under roughly four sources) but can blur attribution once the combined text gets long.learningA product manager has four analyst reports and a dozen saved articles on whether on-device AI or cloud AI will dominate enterprise deployment by 2027, needs to present a one-page synthesis to the leadership team by Friday afternoon, and does not have time to read everything closely enough to defend a recommendation under questioning in the room.

Le Synthétiseur de Recherche : Transformez des Sources Dispersées en une Carte des Accords et des Contradictions

Un prompt IA structuré qui lit plusieurs sources sur la même question et produit une carte de là où elles s'accordent réellement, là où elles se contredisent, et ce qu'aucune source n'aborde — plutôt qu'un résumé lissé qui masque les désaccords.

10/16/2026

decision-makinglearning

Claude Sonnet 5 (also works well with GPT-5.4 and Gemini 2.5 Pro for structured technical writing)Outils de DéveloppementYou're a platform engineer at a payments company shipping a breaking webhook schema change that will affect roughly 400 external integrations, and you need to publish a changelog entry and migration guide before the sprint ends today.

Le rédacteur de changelog pour changements d’API cassants

Un prompt structuré qui transforme un changement d’API cassant en une entrée de changelog claire et un guide de migration étape par étape, en signalant les risques d’échec silencieux avant qu’ils n’atteignent votre file de support.

10/15/2026

developer-experiencetechnical-writing

Works best with Claude Sonnet 5 or Opus-class models — both reliably follow the "do not fabricate missing fields" constraint and flag gaps instead of inventing plausible-sounding compliance language. GPT-6-class models perform nearly as well. Avoid smaller distilled or local models for this one: they are more likely to smooth over missing information with confident-sounding filler, which defeats the entire purpose of an audit record.Intelligence ArtificielleAn HR coordinator used an AI resume-screening tool to auto-reject 40 candidates for a mid-level engineering role on a Tuesday afternoon. On Friday, legal receives a demand letter from a rejected candidate's attorney alleging discriminatory screening, and compliance needs a complete audit-trail record for this specific candidate within 48 hours — including what data went into the AI tool, what it recommended, and who signed off — before legal can respond.

Le Rédacteur de Piste d'Audit des Décisions IA : Transformer une Décision Assistée par IA en Dossier de Conformité Défendable

Un prompt IA structuré qui transforme une seule décision assistée par IA — un filtrage de CV, une décision de crédit, une action de modération de contenu — en une entrée de piste d'audit factuelle conçue pour résister à une demande réglementaire ou à une procédure judiciaire, en signalant les informations manquantes plutôt qu'en les inventant.

10/14/2026

complianceAI governance

Claude Sonnet 5 or GPT-6 — both handle structured multi-column triage well. Needs a model that can hold 15-30 test rows in a table without truncating or re-summarizing them, so avoid smaller distilled models (e.g. Haiku-class) for large test suites — split into batches of about 15 tests at a time if using one.Outils de DéveloppementYour team's CI dashboard has quietly accumulated 18 tests tagged "flaky" over the past two months. Nobody owns the list, and new hires are simply told to click re-run when one of them turns red. On Thursday afternoon, ahead of Friday's release freeze, the QA lead exports the flaky-test tracker — test names, failure rates, and the last few stack traces for each — and runs it through this prompt to get a sprint-ready triage: which tests to quarantine today, which ones trace back to one shared root cause, and which genuinely need another engineer's eyes before anyone touches them.

Le Trieur de Tests Instables : Transformer des Tests CI à Échecs Intermittents en un Plan de Quarantaine Classé par Cause Racine

Un prompt IA structuré qui transforme un tas de tests CI à échecs intermittents en un tableau de triage par cause racine, une liste de mise en quarantaine pour le jour même, et un plan de sprint — pensé pour les équipes lassées de cliquer par réflexe sur « relancer ».

10/14/2026

testingdeveloper tools