Overview

This repository presents pstack, an opinionated agent skill stack that significantly enhances agent outcomes, meticulously ported and adapted for a wide array of AI coding environments. It empowers agents with robust workflows and rigorous verification mechanisms, translating the powerful primitives of Cursor into compatible formats for Claude Code, Codex, Copilot, Pi, Muse, Gemini, Prime Agent, and OpenCode. At its core, pstack intelligently routes tasks to poteto-mode, ensuring your code remains concise, simple, and thoroughly verified.

Key Features

• Multi-Agent Harness Support: Seamlessly integrates with Claude Code, Codex, GitHub Copilot, Pi, Muse, Gemini, Prime Agent, and OpenCode, offering a unified approach to agent workflow management.
• Automated Workflow Routing: Leverages poteto-mode and intelligent routing hooks to automatically select and execute the most appropriate workflow for any given task.
• Formal Verification Integration: Features a /create-formal-verification-skill that identifies critical code properties, verifies them with suitable tools, and integrates the results as project skills for future agent runs.
• Runtime Translation for Claude Code: Pre-translates Cursor skills for Claude Code, optimizing token usage and ensuring accurate mapping at runtime, unlike upstream approaches.
• Extensive Installation Options: Provides clear, step-by-step installation instructions tailored for each supported agent harness, including terminal commands and marketplace integrations.
• Flexible Configuration: Allows customization of model defaults, reasoning effort per role, and automatic routing behavior through the setup-pstack command.
• Comprehensive Playbooks: Offers a rich set of playbooks for debugging, planning, feature development, refactoring, performance optimization, prototyping, PR maintenance, shipping, and extended project work.
• Focus on Verification: Employs skills like how, why, interrogate, tests, and measurements to ensure fixes are not only implemented but also validated against failing and passing evidence.

Typical Use Cases

• Bug Fixing with Verification: Automatically reproduce, diagnose, fix, and verify bugs by leveraging poteto-mode and integrated verification tools, ensuring robust solutions.
• Code Quality Improvement: Maintain concise, simple, and verified code through automated workflow selection and rigorous testing, even for concurrency bugs and hard-to-reach invariants.
• Agent Workflow Enhancement: Elevate the performance and reliability of agents across various platforms by providing a standardized and powerful skill stack.
• Complex Task Management: Handle intricate development tasks like planning, feature implementation, and refactoring with pre-defined and customizable agent playbooks.
• Accelerated Development Cycles: Speed up development by delegating tasks to intelligent agents that can autonomously plan, execute, and verify code changes.
• Formal Verification of Critical Code: Proactively identify and formally verify critical properties in your codebase to prevent elusive concurrency bugs and ensure system integrity.