More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Metaprompter

    1

    Systematically engineer production-grade prompts for agents, API pipelines, and fresh LLM sessions.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Metaprompter

    Metaprompter

    Example session with this skill installed

    Write a prompt for a fresh Claude Code session to audit my repo for unused React components and delete them safely. I want it to stop before deleting anything.

    • Read your context and instructions
    • Compiled the metaprompter

    I have generated a multi-step agent prompt for a repository audit and cleanup.

    Assumptions

    • The repository uses a standard src/ directory structure.
    • grep or rg is available in the environment to check component usage.
    • You want the agent to check for imports in .js, .jsx, .ts, and .tsx files.

    How to use this

    1. Open a new Claude Code session in your repository root.
    2. Paste the prompt below.
    3. Review the proposed deletion list when the agent hits the stop point.
    
    # Task: Unused React Component Audit and Cleanup
    
    You are an expert developer tasked with identifying and removing dead code. You will audit the current repository to find React components that are defined but never imported or used elsewhere in the project.
    
    ## Context and Scope
    - Target Directory: `src/`
    - File Extensions: `.js`, `.jsx`, `.ts`, `.tsx`
    - Exclude: Test files (`*.test.*`, `*.spec.*`), Storybook files (`*.stories.*`), and `index.ts/js` barrels.
    
    ## Instructions
    1. **Inventory**: Map all component files in `src/`. A component file is any file that exports a React component.
    2.
    
    **Usage Check**: For each identified component, search the entire `src/` directory (excluding the file itself) for import statements or dynamic references to that component.
    3.
    
    **Verification**: If a component appears unused, verify it is not exported as part of a public API if this is a library (check `package.json` for entry points).
    4.
    
    **Reporting**: Create a Markdown table of all components identified as "Unused". Include the file path and the last modified date.
    
    ## Guardrails and Stop Points
    -
    
    **STOP**: After generating the report table, you must stop and ask for explicit user approval before proceeding to deletion.
    - **READ-ONLY**: Do not modify any files until the user provides the "APPROVED" keyword.
    - **BACKUP**: If approved, move files to a temporary `.trash/` directory instead of immediate hard deletion.
    
    ## Reasoning Requirements
    Before listing a component as unused, explain your reasoning in a `<scratchpad>` block, noting the search strings used to verify its absence.
    

    Next steps

    • Ensure you have a clean git state before running so you can revert changes.
    • Provide the agent with any specific directories to ignore if they contain experimental code.

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

    About this skill

    High-quality output from an LLM depends entirely on the instructions it receives. Generic prompts lead to generic results, while manual prompt engineering is a time-consuming cycle of trial and error. This skill automates the creation of production-grade instructions for other agents, API pipelines, or fresh sessions.

    What it does

    • Shape selection identifies whether the task requires a multi-step agent session for complex work or a single-turn template for high-volume API processing.
    • Context gathering extracts specific IDs, file paths, URLs, and project constraints from your environment to ensure the prompt is grounded in reality.
    • Protocol design builds in explicit guardrails, stop points for human approval, and named fallbacks for tool failures.
    • Formatting precision enforces exact output schemas, XML tagging for long-form data, and scratchpad reasoning steps to improve logic.

    How it works

    1. Analyze the request to determine if the target is a "stranger" session (like Claude Code) or a structured pipeline.
    2. Scan available resources including project docs, repos, and live APIs to find the specific facts the target model will need.
    3. Draft the instruction set using a professional, technical tone that prioritizes reasoning before verdicts and includes concrete stop points.
    4. Validate and deliver the prompt in a clean, copy-pasteable format with a short list of assumptions and usage steps.

    Frameworks & tools

    Works with any LLM environment including Claude Code, Cursor, GitHub Copilot, and custom agentic pipelines via API.

    Why this beats prompting it yourself

    Most manual prompts lack the necessary boundaries and context to run autonomously without hallucinating or getting stuck. This skill applies a systematic engineering framework to include error handling and resource references that developers often overlook.

    Use cases

    • Generate a system prompt for a new agent to perform codebase refactoring.
    • Create a reusable template for an API pipeline to classify support tickets.
    • Hardened an existing prompt by adding approval steps and tool fallbacks.
    • Write a one-shot prompt for a fresh session to conduct technical research.

    Known limitations

    Does not support generating secrets, keys, or tokens. Requires the user to provide or point to the necessary project context for high-accuracy results.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Prompt & Skill Engineering.

    Does not generate secrets/keys. Effectiveness scales with the amount of project context (docs, repos, APIs) provided or accessible during the gathering phase.

    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 buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 13 days ago

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed13 days ago

    What's inside

    Frequently Asked Questions

    Metaprompter automates the creation of production-grade instructions by analyzing your specific project context, enforcing logic-heavy formatting like scratchpads, and building in error-handling protocols that manual prompts often miss.

    Yes, it is environment-agnostic; it generates optimized instructions for any platform, including Claude Code, Cursor, GitHub Copilot, and custom API-driven pipelines.

    The skill works by scanning your provided project docs, repos, and live APIs to identify the exact facts and constraints the target model needs, ensuring the output is grounded rather than generic.

    You receive the full Metaprompter skill set, including the automated shape selection logic, context extraction framework, and professional protocol design templates for immediate use.

    Metaprompter is designed to identify whether your task requires a multi-step agent flow for complex reasoning or a single-turn template for high-volume API processing, selecting the optimal 'shape' for you.

    While it can generate complex logic and guardrails, it cannot generate or store sensitive data like API keys, secrets, or authentication tokens.

    New in Prompt & Skill Engineering