Agent Task Prompt Generator

Generate structured task prompts for coding agents — Claude Code, Cursor, Codex, Windsurf, Gemini CLI, and more. Fill in the task, requirements, constraints, and verification steps — get a paste-ready prompt, a Markdown spec, and a verification checklist.

Target Agent

Task Details

Task title

Codebase context

Describe the project so the agent understands the environment

Goal

Requirements & Constraints

Requirements

Constraints

Files likely affected

Do NOT touch

Verification & Completion

Implementation notes

Optional hints for the agent

Verification commands

Manual test checklist

Final response requirements

Commit message

Optional

// Fill in task title and goal to generate…

Nothing is uploaded.

How to write better coding agent prompts

🏗

Always provide codebase context

Describe the project, tech stack, and relevant architecture. Without context the agent guesses and often guesses wrong.

🎯

State the goal clearly

"Ensure all tool counts derive from tools.ts, not hardcoded values." One sentence that captures the outcome — not the steps.

📋

List requirements explicitly

Number them. Ambiguous requirements lead to partial implementations. Each requirement should be independently verifiable.

🚫

Name files to avoid

dist/, .env, generated files. Agents sometimes edit build output or configuration files that should not be touched.

✅

Include verification commands

pnpm build, tsc --noEmit, pytest. The agent should run these after implementing and confirm they pass.

📊

Define the final report

Tell the agent what to report when done. "List every changed file and confirm all checks pass" prevents vague completions.

Supported coding agents

Claude CodeOpenAI CodexCursorWindsurfGemini CLIGeneric Coding Agents

Generated prompts are starter templates — always review and adjust for your specific agent's behavior and capabilities.

Privacy: This tool runs in your browser. Your task details are never uploaded or stored.

Frequently Asked Questions

What is a coding agent task prompt?

A coding agent task prompt is a structured message you give to a coding agent (like Claude Code, Cursor, or Codex) telling it exactly what to do. Unlike a casual chat message, a good task prompt includes codebase context, explicit requirements, constraints, which files to touch and which to avoid, verification commands, and a final report requirement.

Why does context matter so much?

Coding agents work best when they understand the project structure, tech stack, and current state. Without context, the agent makes assumptions that may be wrong — creating files in wrong locations, using incompatible patterns, or missing project conventions.

What is the "Do NOT touch" section for?

It explicitly lists files and directories the agent must not modify. Common entries: dist/ (build output), .env (secrets), pnpm-lock.yaml (managed by package manager), and database migration files.

What is the Compact Prompt output?

The Compact Prompt distils the full prompt into a shorter version — useful for agents with smaller context windows or when you want a quick summary version without the detailed breakdown.

What should go in the Final Report Requirements?

Tell the agent exactly what to report when done. Example: "List every file changed, confirm all verification commands pass, and confirm no hardcoded values remain in the UI." This ensures you get actionable output, not just "done".

Is my data uploaded anywhere?

No. This tool runs entirely in your browser. Nothing is sent to any server.

Agent & Skill Generators