system-prompt-curator

Expert system for creating and refining production-grade system prompts for autonomous coding agents and multi-agent orchestration.

This plugin is the canonical reference for writing high-quality agent identities, workflows, and completion criteria that work reliably across Claude Code, Cursor, Codex, OpenCode, Kiro, Devin-style agents, and other modern coding agents.

What it does

• Generates new system prompts from role descriptions
• Dramatically improves existing prompts by detecting research-backed anti-patterns
• Provides specialized templates for common agent roles (GitHub issue → PR, research, review, orchestration)
• Enforces the 10 core principles proven across top-performing agents (SWE-agent, OpenHands, Claude Code, Cursor, etc.)
• Recommends harness-level reinforcements that belong in code rather than the prompt

Installation

agentsys install system-prompt-curator

Usage

/system-prompt-curator "senior software engineer that resolves GitHub issues end-to-end"

/system-prompt-curator --improve path/to/existing-prompt.md --for-orchestrator

/system-prompt-curator --improve path/to/prompt.md --minimal

Related Plugins

• skill-curator — sibling plugin for writing excellent SKILL.md files
• enhance — broader prompt/plugin/agent improvement
• karpathy-guidelines — foundational behavioral rules often included in prompts

Philosophy

The best agents are not the ones with the cleverest tricks in the prompt — they are the ones with:

• Clear identity
• Enforced phased workflow
• Concrete completion criteria that require evidence
• Proper error recovery guidance
• Tools declared upfront

This curator exists to make those patterns easy to apply consistently.