agentic-system-prompts

The agentic-system-prompts library documents system prompts, tool definitions, and design patterns from production AI coding agents including Claude Code, Gemini CLI, Cline, Aider, Roo Code, Zed, and Codex CLI. Each agent directory includes complete prompt templates, detailed tool API specifications, and source attribution to help developers understand how professional agentic systems structure instructions and capabilities. This collection provides research value for prompt engineering patterns, tool design approaches, safety measures, and user experience workflows across different agent implementations. Operations and development teams can use these documented prompts as reference material or starting points for building consistent, reliable AI agent configurations.

[{"step":1,"action":"Select a system prompt from the curated collection (e.g., `refactor-service-component`, `debug-performance-issue`, `generate-api-docs`).","tip":"Use prompts tagged for your tech stack (e.g., 'Spring Boot', 'React Native') to ensure compatibility."},{"step":2,"action":"Customize the prompt template by replacing [PLACEHOLDERS] with your task details (e.g., [TASK_DESCRIPTION], [CONSTRAINTS]).","tip":"For dynamic inputs, use Jinja templating (e.g., `{{ project_name }}` → `auth-service`)."},{"step":3,"action":"Deploy the prompt to your AI agent (e.g., Claude, Cursor, or custom agent). Ensure the agent has access to the required tools (e.g., `dependency-graph-tool`, `maven-wrapper`).","tip":"Test the agent in a sandbox environment first to validate tool outputs and constraints."},{"step":4,"action":"Monitor the agent’s execution log. Review outputs for accuracy, and use the documented steps to verify changes or debug issues.","tip":"For complex tasks, break the workflow into smaller subtasks and use separate prompts for each phase."},{"step":5,"action":"Integrate the agent’s outputs into your workflow (e.g., auto-generate PR descriptions, update documentation, or trigger CI/CD pipelines).","tip":"Use the agent’s structured output (e.g., JSON, Markdown) to automate downstream processes like ticket updates or Slack notifications."}]