Scoped edits
Prompts that limit changes to named files and forbid unrelated cleanup.
24 curated prompts
Cursor Prompts
Coding assistants inside an IDE make many small edits, and the failure mode is usually scope creep rather than bad syntax. These prompts constrain the change surface, require a plan before edits, and keep every modification reviewable.
Cursor promptsAI coding assistant promptsIDE agent prompts
Plan before code
Templates requiring a short plan and a rollback note before any diff.
Tests and review readiness
Prompts that generate tests and explain how to verify the change.
Featured AI prompt templates
Copy-ready prompts selected from this topic cluster.
TextCoding
Coding Agent System Prompt
A system prompt for a software engineering assistant embedded in the command-line development environment, guiding AI to complete programming tasks such as coding, debugging, refactoring, and test management. Emphasizes safe operations, cautious execution of changes, preference for dedicated tools, and requires output in GitHub-flavored Markdown format.
TextCoding
Terraform/OpenTofu Infrastructure as Code Specialist
Manages infrastructure code as production-grade software, emphasizing versioning, testing, and rollback capabilities. Enforces strict response contracts with pre-diagnosis of failure modes.
TextCoding
Act as a Senior Codebase Auditor & Tech Debt Refactorer
Senior Codebase Auditor & Tech Debt Refactorer, focusing on code quality, maintainability, and technical debt auditing.
TextCoding
PCB/EDA Design Architect
This prompt defines a senior PCB/EDA design architect responsible for end-to-end review of electronic designs, covering schematic, PCB layout, signal integrity, power integrity, EMC pre-compliance, SPICE simulation, and DFM analysis, delivering structured engineering reports.
CodeCoding
Embedded Firmware Engineer Prompt
A comprehensive prompt for senior embedded firmware engineers, covering production-grade firmware development on resource-constrained MCUs, RTOS task design, communication protocols, memory safety, and cross-platform portability.
TextCoding
Refactoring Coach
This prompt guides the model to act as a senior refactoring specialist who helps developers systematically improve maintainable but problematic code while preserving all existing behavior through safe, incremental transformations.
CodeCoding
Andrej Karpathy Coding Guidelines: Behavioral Rules to Reduce Common LLM Coding Mistakes
A set of behavioral guidelines designed to minimize common errors in LLM-generated code, emphasizing caution, simplicity, surgical changes, and goal-driven execution.
TextCoding
Comprehensive Codebase Auditor
Exhaustively analyze entire codebase to identify, classify, and prioritize all verifiable bugs, security vulnerabilities, and critical issues.
CodeCoding
Anti-AI-Slop Design Architect
An opinionated frontend design skill prompt that enforces structural variety, honest copy, locked design tokens, and a hard mobile-responsiveness floor, then verifies output against an anti-slop gate battery.
TextCoding
Self-Distillation Code Generation Strategist
Evaluates whether self-distillation is the optimal next step for a given code generation task based on diagnostic pass@k metrics, designs experimental pipelines if applicable, and refuses recommendations when outside its operating envelope.
TextCoding
Test Strategy Architect: Building a Sustainable Testing Framework
Design and implement a layered, maintainable testing strategy for complex software systems to ensure high coverage, fast feedback, and long-term sustainability.
TextCoding
Senior Python Code Auditor (Spanish Output)
Act as a senior software architect and Python expert to perform a comprehensive code audit and complete refactoring of the provided script, with output in Spanish.
Why do IDE agent prompts need scope limits?
Without them the assistant rewrites unrelated code, which makes review difficult and increases regression risk.
Should I paste whole files into these prompts?
Give the smallest context that reproduces the task. Include interfaces and call sites instead of entire repositories.
How do I make the assistant explain its reasoning without wasting turns?
Ask for a brief plan and a list of assumptions, not a full chain of thought.