Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Prompt Generator & Optimizer (PT-BR + EN)

    3

    Turns a one-line request into a compact and a full copy-ready prompt, or diagnoses and rewrites the prompt you paste.

    Free

    2 uses5.0 (1 review)
    Prompt Generator & Optimizer (PT-BR + EN)

    Prompt Generator & Optimizer (PT-BR + EN)

    Example session with this skill installed

    write a prompt to review a pull request in our Python API before merge

    • Read your context and instructions
    • Compiled the prompt generator

    Category: Code / Technical · Assumptions: Python 3.11+, FastAPI; output = prioritized review comments

    Compact prompt

    You are a senior Python engineer reviewing a pull request for a production API.
    Review {{DIFF}} for bugs, security issues, missing tests and unclear naming.
    List findings by severity (blocker / major / minor). For each: file:line, the problem, a concrete fix.
    End with one line: merge / merge after fixes / do not merge.
    

    Full prompt

    <role>Senior Python engineer, expert in FastAPI, async code and API security.</role>
    <context>Pull request to {{BRANCH}} in a production API. The reviewer must catch issues before merge, not restyle the code.</context>
    <task>Review the pull request in {{DIFF}} and decide whether it can be merged.</task>
    <rules>
    - Check correctness, input validation, authentication/authorization, error handling, test coverage and naming.
    - Think step by step through each changed file before concluding.
    - Report only real issues; skip pure style preferences unless they hurt readability.
    - Mark anything you cannot verify from the diff as "needs context".
    </rules>
    <format>
    Findings grouped by severity (blocker / major / minor).
    For each: file:line · problem · suggested fix (code if short).
    Final line: merge / merge after fixes / do not merge, with one-sentence reason.
    </format>
    

    Why it works: severity order and a concrete fix per finding turn the review into an action list instead of a list of opinions.
    To refine: add your team's checklist (e.g., logging, migrations); or ask for a summary comment ready to paste on the PR.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Turn a one-line request into a strong, copy-ready prompt — without wasting tokens.

    What it does

    • Classifies your request and applies a domain module: Brazilian law, civil-service exams (Cebraspe, FGV, FCC), management, finance, study, code, content, fitness, research and creative.
    • Asks at most 2 questions, only when something critical is missing. Otherwise it states its assumptions as {{VARIABLES}}.
    • Delivers two versions: a compact prompt (≤60% of the tokens) and a full one with role, context, task, rules and output format.
    • Optimizes existing prompts: paste yours and get a 3-point diagnosis (missing / excess / ambiguity) plus a rewritten version.

    Output format

    Category and assumptions → compact prompt → full prompt → one line on why it works → up to 2 refinement ideas.

    Languages

    Portuguese (Brazil) first; writes the prompt in the same language as your request.

    Triggers

    "gera um prompt", "melhora este prompt", "/prompt", "write a prompt", "improve this prompt".

    Generates prompts only; it does not execute them unless asked. Asks at most two clarifying questions. Image prompts skip XML syntax. Primary focus is PT-BR and Brazilian exam boards (Cebraspe, FGV, FCC). Legal citations produced by the generated prompts must be checked against current law and case law.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    5.0
    1 review
    5
    1
    4
    0
    3
    0
    2
    0
    1
    0

    This skill has been used 2 times.

    Trust & safety

    Security scan

    Security scan passed 13 days ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed13 days ago

    What's inside

    Frequently Asked Questions

    It detects the domain of your request and adds that domain's must-haves to the prompt. For Brazilian law: facts, legal question, grounds (Constitution, statutes, STF/STJ case law) and a check for current validity. For civil-service exams: exam board, question style and the topic from the official notice. A generic prompt usually leaves these out.

    By default it writes prompts for Claude. If you name another target (ChatGPT, Gemini, Midjourney), it adapts the syntax, for example dropping XML tags for image generators. The skill itself runs in any SKILL.md-compatible agent.

    You get two versions: a full prompt with every block, and a compact one that targets at most 60% of its tokens while keeping every critical constraint. Use the compact one for frequent or API use; use the full one when the task is complex.

    One SKILL.md with the full workflow, domain modules and output format, plus a complete input/output example in reference/exemplo.md. No scripts, no dependencies, no permissions.

    If something that changes the result is missing (jurisdiction, exam board, audience, deadline), it asks up to two questions in a single message. Otherwise it fills sensible defaults and marks them as {{VARIABLES}} so you can edit them.

    No. By default it only generates the prompt; you then run it in your AI tool. If you want both, ask "gera e já executa" (generate and run) and it will do the task right after writing the prompt.

    New in Prompt & Skill Engineering