text
Prompt de configuration et widgets AI Desk Card
Prompt pret a utiliser pour installer AI Desk Card, flasher le M5Paper, configurer le Wi-Fi et pousser des widgets utiles.
Claude CodeCodexCursorAiderGemini CLI
110
AI Prompts
10 prompts available
text
Agent Safety Red-Team Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to stress-test an agent workflow for prompt injection, data leakage, tool misuse, and unsafe autonomy, with evidence checks, output structure, and source discipline.
claudegptgemini
80
text
SEO Content Refresh Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to rewrite old AI directory content with current sources, media, FAQs, alternatives, and snippet-ready sections, with evidence checks, output structure, and source discipline.
gptclaudegemini
50
text
RAG Source Calibration Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to audit documents and retrieval outputs for source quality, freshness, coverage, and hallucination risk, with evidence checks, output structure, and source discipline.
gptclaudegemini
100
text
AI Agent Builder Spec Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to describe an agent in enough detail to choose tools, permissions, memory, evals, and human approval points, with evidence checks, output structure, and source discipline.
gptclaudegemini
110
text
Prompt Versioning and Eval Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to turn a production prompt change into a testable version with rollback notes and eval cases, with evidence checks, output structure, and source discipline.
gptclaudegemini
70
text
Vibe Coding PRD-to-Tasks Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to translate an idea into a scoped implementation plan that coding agents can execute safely, with evidence checks, output structure, and source discipline.
claudegptgemini
100
text
Product Hunt Launch Intelligence Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to convert a Product Hunt launch page into positioning, objections, audience, and content angles, with evidence checks, output structure, and source discipline.
gptclaudegemini
90
text
GitHub Trending Scout Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to scan repositories, star velocity, maintainer quality, and product relevance before adding a project, with evidence checks, output structure, and source discipline.
gptclaudegeminigrok
100
text
Deep Research Brief Prompt
Contenu pratique pour les workflows IA en 2026 : A copy-ready prompt to turn a vague research question into a sourced, decision-ready brief, with evidence checks, output structure, and source discipline.
gptclaudegemini
70
Prompt engineering in 2026
The useful prompt is no longer a clever sentence. It is a tested instruction interface.
A prompt library should help users control output, not just collect catchy templates. The strongest prompts now define the task, the context boundary, the desired format, examples, and the evaluation rule. That matters because the same request behaves differently across chat models, reasoning models, image generators, and agent runtimes.
What changed
Clarify capability boundaries before adoption.
From prompt to contract
A good prompt states what success looks like
OpenAI’s guidance still starts with practical basics: put instructions first, separate reference text clearly, specify output format, and use examples when structure matters. For directory users, that means the best prompt cards should be judged by repeatability, not by how impressive the first draft sounds.
Context is the real bottleneck
The prompt is only one part of the model’s working memory
LangChain frames context engineering as selecting the right instructions, knowledge, tools, memories, and prior messages for the next model call. This is why modern prompts should include context slots and source boundaries instead of pretending one universal mega-prompt works for every case.
Formats beat vibes
Structured output makes prompts easier to test
Prompts that ask for tables, JSON, step lists, grading rubrics, or before/after comparisons are easier to review and reuse. They also make it clearer when a model ignored the task, invented facts, or produced content that cannot be used downstream.
Popular starting points
Find AI prompts by model, use case, and output type
Browse prompt templates by what you want to create: Gemini AI prompts, AI prompt generator workflows, AI prompt examples, AI photo prompts, photo-editing prompts, and AI video prompt formats.
gemini ai prompt
Template discoveryUseful for users who want Gemini-specific examples, copy-paste structures, photo prompts, and style prompts without sorting through generic ChatGPT-only templates.
ai prompt generator
Tool-assisted creationStrong commercial value. Users want a repeatable way to generate prompts, not just a static list of phrases.
ai prompt examples
Learning and comparisonLow competition in the sheet and very high CPC. Add examples, before/after structure, and use-case explanations.
Explore AI prompt use cases
Filter by the job to be done.
Use the page to catch four intent types: broad AI prompt discovery, generator workflow searches, Gemini copy-paste searches, and visual prompt searches.
Prompt generator intent
These users want a tool-like workflow. The copy should explain how prompts are generated, refined, evaluated, and adapted across models.
ai prompt generatorai prompt image generatorai new prompt
Gemini prompt intent
These users want ready-to-use Gemini prompts, especially for photos and viral styles. Keep the exact phrases but avoid turning the page into a thin copy-paste dump.
gemini ai promptai gemini promptgemini ai prompt copy pastegemini ai prompt photo
Image and video prompt intent
These queries need visual production language: subject, camera, lighting, motion, negative prompt, revision checks, and platform notes.
ai photo promptai prompt imageai prompt for photo editingai video prompt
How to choose prompt types
Compare by capability, cost, and risk.
Use the prompt type that matches the risk of the task. A social caption prompt and a legal research prompt should not be evaluated with the same standard.
Creative prompts
- Best for ideation, images, video, ads, and naming
- Need style, audience, medium, constraints, and variation count
- Quality metric: range of useful options
Operational prompts
- Best for reports, summaries, extraction, support replies, and SEO briefs
- Need source text, schema, tone, and rejection rules
- Quality metric: consistency across repeated runs
Agent prompts
- Best for multi-step work with tools or files
- Need goal, available tools, stop conditions, and approval rules
- Quality metric: correct actions, not just polished answers
Practical selection checklist
Fit the tool into a real workflow.
Prefer prompts with variables such as {{audience}}, {{source_text}}, {{format}}, and {{constraints}}.
For factual work, require the model to separate source-supported claims from assumptions.
For image and video prompts, look for camera, composition, lighting, motion, brand-safety, and avoid-list fields.
For business workflows, choose prompts that include acceptance criteria and a revision loop.
Related prompt topics
Use these topics to move from broad AI prompt discovery into specific examples, image prompts, Gemini prompts, and video prompt workflows.
ai promptai prompt generatorai prompt examplesgemini ai prompt copy pasteai image editor with prompt freeai prompt for photo editingai video promptgemini ai prompt retro style
Prompt FAQ
Common questions and selection answers.
Are prompt marketplaces still useful?
Yes, if they curate tested templates instead of decorative text. A prompt marketplace is useful when it saves users from rebuilding task structure, examples, and output rules from scratch.
What is the fastest way to improve a weak prompt?
Add the missing context and output contract: who the output is for, what source material it may use, what format it must return, and what it should do when the input is incomplete.
Should prompt pages mention specific models?
Yes, when model behavior matters. Reasoning models, chat models, and image/video models respond differently, so platform tags help users avoid copying the wrong pattern into the wrong tool.