gemini-deep-research
Skill1 starsChanged 5 months agoMITView on GitHub
Lance une Deep Research sur gemini.google.com via Chrome MCP (extension Claude for Chrome). Attend la complétion (5-15 min), extrait le rapport complet, sauvegarde en Markdown. Use when the user wants a deep research done on a topic via Gemini Advanced without manual copy-paste.
8 other projects use a version of this.
Adopted by
Most recently taken up by
- OpenLAIR/dr-clawtheir own version, first seen Nov 8, 2026
- amanuelmamo7/Skillstheir own version, first seen Nov 7, 2026
- yigityildiz0/evidence-research-skillstheir own version, first seen Oct 25, 2026
- iberi22/skillstheir own version, first seen Oct 15, 2026
- VisionaireLabs/Visionairetheir own version, first seen Sep 30, 2026
- jeffcryder/skillstheir own version, first seen Aug 30, 2026
Other versions8 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 8
- OpenLAIR/dr-claw★ 1.2ksame name
- zcaceres/skills★ 7same name
- VisionaireLabs/Visionaire★ 4same name
- amanuelmamo7/Skills★ 1same name
- hiapplyco/claude-skills★ 0same name
- iberi22/skills★ 0same name
- jeffcryder/skills★ 0same name
- yigityildiz0/evidence-research-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- anti-ai-detect-writingSkill
- audit-short-1v1Skill
- audit-short-ffaSkill
- awesome-design-mdSkill
- claude-councilSkill
- deep-researchSkill
- design-mdSkill
- design-audit-paris-nuitSkill
- gemini-spriteSkill
- oracle-ampere-a1-setupSkill
- smart-routerSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 143
If incomplete: : use INCOMPLETE: <missing / cause / needs> — never claim "done" with empty output.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
gemini-deep-research
Automatisation complète d'une Deep Research Gemini Advanced depuis Claude Code.
Prérequis (déjà OK côté {{USER_NAME}})
- Extension Claude for Chrome installée (v1.0.68 confirmée, ID
fcoeoabgfenejglbffodgkkbkcdhcgfn) - Claude Desktop ouvert avec extension Chrome connectée
- Session
gemini.google.comloggée sur compte Gemini Advanced / Google One AI Premium - MCP
mcp__Control_Chrome__*oumcp__Claude_in_Chrome__*dispo dans la session
Usage
/gemini-deep-research <sujet de recherche détaillé>Workflow
1. Vérifier Chrome MCP
Si aucun tool mcp__Control_Chrome__* ni mcp__Claude_in_Chrome__* disponible : ping {{USER_NAME}} pour rouvrir Claude Desktop + extension Chrome, puis relancer Claude Code.
2. Ouvrir Gemini Advanced
Appel mcp__Control_Chrome__open_url → https://gemini.google.com/app
Attendre 3 sec (chargement SPA).
3. Activer le mode Deep Research
Via mcp__Control_Chrome__execute_javascript, stratégies de recherche du bouton dans l'ordre :
- Par aria-label :
[aria-label*="Deep Research" i]ou[aria-label*="Recherche approfondie" i] - Par text content sur
buttonet[role="button"] - Via menu "tools" / "outils" qui ouvre un sous-menu contenant l'option
Cliquer dès trouvé.
4. Coller le prompt et soumettre
JS à exécuter :
- Trouver éditeur :
rich-textarea .ql-editor,[contenteditable="true"], ou fallbacktextarea - Focus + clear content (textContent pour contenteditable, value="" pour textarea)
- Insérer le prompt via
document.execCommand('insertText', false, prompt) - Dispatcher un Event
inputbubbles:true - Trouver bouton Send via aria-label (
Send,Envoyer,Submit) - Click
5. Démarrer la recherche (si plan intermédiaire)
Gemini propose souvent un plan de recherche à valider. Poll 20×3sec pour trouver un bouton dont le text content matche /start research|lancer la recherche|démarrer/i. Clic dès trouvé. Si non trouvé en 60sec, considérer que la recherche a démarré directement.
6. Polling complétion
Toutes les 30 sec (max 20 itérations = 10 min), exec JS pour détecter la fin :
- Absence de progress bar
[role="progressbar"] - Présence d'un bouton export / share avec aria-label matchant
/export|share|partager/i
Si > 10 min, ping {{USER_NAME}} avec option de continuer ou arrêter.
7. Extraire le résultat
Charger et exécuter le fichier extract_result.js (livré avec le skill) via mcp__Control_Chrome__execute_javascript. Retourne un objet avec :
successmarkdown(conversion HTML→MD basique)text(plain)word_countlinks(sources extraites)
8. Sauvegarder
Écrire dans :
~/Desktop/workspace/08_VEILLE/deep-research/deep-research-<slug>-<YYYY-MM-DD>.mdFormat :
---
source: Gemini Advanced Deep Research
date: 2026-04-20
topic: "<sujet original>"
word_count: <nombre>
duration_minutes: <durée observée>
sources_count: <nombre de liens>
---
# <titre>
<contenu markdown extrait>
## Sources
- [Texte](url)
- ...9. Notifier {{USER_NAME}}
- Chemin du fichier
- Nombre de mots + nombre de sources
- Durée réelle
- TL;DR 3-bullets extrait des premiers paragraphes
Fallback
Si Chrome MCP indisponible : proposer Option 1 (skill /deep-research basé sur WebSearch parallèles, gratuit, 3-5 min).
Robustesse
- Extension Claude for Chrome = vraie intégration navigateur, pas de détection anti-bot
- Sélecteurs multi-fallback (aria-label + text content + menu navigation)
- Timeout global 20 min ; au-delà, extraction partielle + notification
Exemples de prompts
Tech :
Évolution prix LLM APIs sept 2025 → avr 2026 : Anthropic, OpenAI, Mistral, Google.
Prix input/output/M tokens, tokenizer changes, DPA, hébergement EU.
Tableau final + reco freelance IA FR B2B souverain.Marché :
Freelance IA France 2026 : TJM moyen par profil (Dev IA, expert RAG, LLMOps).
Plateformes Malt/Comet/FreelanceRepublik.
Trends missions "Assistant IA custom" vs "RAG pipeline".Prospection :
Recherche approfondie sur <entreprise cible>.
Secteur, CA, dirigeants, outils IA actuels, 10 dernières actus.
3 opportunités mission IA + profil LinkedIn décideur IT.OUTPUT CONTRACT
Respect ~/.claude/rules/output-contract.md. Before claiming the task done:
- Declare the deliverable path explicitly (file or artifact)
- Run testable checks (
test -f,wc,grep,jq, smoke test command — paste output) - Final response format :
LIVRABLE: <one-liner>
FICHIER: <absolute path> (<size/lines>)
CHECKS PASS: X/Y (paste failing ones)
NEXT {{USER_NAME}} ACTION: <what he should do with the output>- If incomplete : use
INCOMPLETE: <missing / cause / needs>— never claim "done" with empty output.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

