Prompts reutilizáveis: transformar padrões em templates para várias ferramentas de IA (Ae)

Role: You are a Prompt Architect helping teams capture and reuse prompt patterns across AI tools (ChatGPT, Claude, Gemini, Copilot, Perplexity).

Goal: Produce a ready-to-use blueprint that enables cross-tool prompt reuse via a template library. Deliver a JSON array of template objects. Each template should include:
- id
- name
- description
- sections: [role/context, task definition, constraints, examples, output_format, tone rules, variables]
- variables: [{ name, type, default }]
- example_input
- example_output
- recommended_tools

Constraints:
- Templates must be tool-agnostic and work across ChatGPT, Claude, Gemini, Copilot, and Perplexity.
- Use placeholders in braces, e.g., {user_name}, {deadline}, {data_source}.
- Provide at least 5 templates.
- Include a high-level plan to implement a browser extension/workspace named 'Ae' that stores, refines, and shares templates via a public marketplace; describe frontend, backend, storage, and a simple data model.
- Output a single valid JSON array of templates. If helpful, you may include an optional final object describing the Ae architecture, but the overall output must remain valid JSON.

Examples to include (3 templates):
1) Summarize document
- id: tmpl_summarize_doc
- name: Summarize document
- description: Produce a concise summary with key points tailored to an audience
- sections: [role/context: "You are a summarization assistant.", task_definition: "Summarize the provided document or text.", constraints: "Limit to X words or Y bullets; preserve factual accuracy; use bullet points unless instructed otherwise.", examples: "Provide a brief, structured summary.", output_format: "Bullet list with 3-7 bullets; optionally include a one-line takeaway.", tone: "Neutral, factual"]
- variables: [{ name: input_text, type: string, default: "" }, { name: max_words, type: integer, default: 150 }, { name: audience, type: string, default: "general" }]
- example_input: "[Paste document here]"
- example_output: "- Key point 1\n- Key point 2\n- Key point 3"
- recommended_tools: ["chatgpt", "claude", "gemini"]

2) Extract entities
- id: tmpl_extract_entities
- name: Extract entities
- description: Identify and classify named entities (persons, organizations, dates, etc.) from the input text
- sections: [role/context: "You are an information extraction assistant.", task_definition: "Extract and classify entities.", constraints: "Use a consistent tagging scheme; output as JSON.", examples: "", output_format: "JSON array of entities with type and text.", tone: "Neutral"]
- variables: [{ name: input_text, type: string, default: "" }]
- example_input: "[Paste text here]"
- example_output: "[{"text": "ACME Corp", "type": "ORG"}, {"text": "2024-05-01", "type": "DATE"}]"
- recommended_tools: ["chatgpt", "perplexity"]

3) Code review assistant
- id: tmpl_code_review
- name: Code review assistant
- description: Provide a structured code review with issues, suggestions, and refactoring steps
- sections: [role/context: "You are a code reviewer.", task_definition: "Review the provided code and output a checklist of issues, suggestions, and refactors.", constraints: "Be specific, provide line-number references, and avoid boilerplate.", examples: "", output_format: "List of issues with severity, then suggested changes.", tone: "Professional, constructive"]
- variables: [{ name: code, type: string, default: "" }, { name: language, type: string, default: "" }]
- example_input: "<code snippet here>"
- example_output: "1. Issue: ... (Line 23)\n   Suggestion: ...\n2. Issue: ..."
- recommended_tools: ["github-copilot", "chatgpt"]

Architecture and Ae plan:
- Provide a high-level architecture sketch for the Ae extension: frontend (prompt editor UI), backend (template engine and marketplace APIs), and storage (templates, versions, user data).
- Include a simple data model (Template, Variant, Tag, Version, User, MarketplaceListing).
- Include basic versioning, discovery, and sharing guidelines.

Finally, analyze and justify which AI tool from the list (chatgpt, claude, deepseek, gemini, copilot, perplexity) is best suited to maintain and evolve this reusable-prompt system and why.