open-canal: Guide to Software Project Specification Methodology and Skill Suite for AI Coding Agents

open-canal is a complete AI software specification methodology that helps AI coding agents like Claude Code and Codex CLI perform systematic software project planning and document management through composable skills, shared standards, and templates. The project is maintained by sakuraine, sourced from GitHub (link: https://github.com/sakuraine/open-canal), and released on May 25, 2026. Its core goal is to solve the problem of AI coding agents lacking systematic specifications in complex projects, convert software engineering best practices into AI-executable skills, and improve the efficiency of project planning and document management.

Background and Motivation: Project Planning Challenges Faced by AI Coding Agents

With the rapid development of AI coding agents like Claude Code and Codex CLI, developers rely on them for auxiliary development. However, requirements documents, design specifications, and other elements in traditional development processes are difficult to be effectively utilized by AI, leading to unsystematic project planning. open-canal emerged to convert software engineering best practices into "skills" executable by AI agents, enabling AI to perform project planning and document management like experienced architects.

Core Concepts and Architecture: Composable Skills and Obsidian Knowledge Base Support

open-canal follows the Agent Skills specification standard, with its core being the decomposition of software specification processes into composable and reusable skill modules. The project centers around the Obsidian knowledge base; after initialization, it creates a .open-canal directory containing standards (standards) and templates (templates) subdirectories, providing infrastructure for specification work.

Core Skill Suite: Seven Workflows Covering the Entire Lifecycle

open-canal provides seven core skills covering the software development lifecycle:

1. framework-init: Initialize the Obsidian specification library, establish directory structure and configuration;
2. requirement-create: Convert functional ideas into structured PRD (including acceptance criteria);
3. version-create/version-add: Manage version iterations and maintain traceability between requirements and versions;
4. design-create: Generate prototype prompts and manage the DESIGN.md design system file;
5. develop-create: Generate technical implementation plans for clients, servers, and databases;
6. test: Automatically generate test plans based on PRD and interface contracts.

Multi-Platform Compatibility: Supports Mainstream AI Coding Agent Tools

open-canal is compatible with multiple mainstream AI coding agents, including Anthropic's Claude Code, OpenAI's Codex CLI, the emerging Crush, and the open-source OpenCode. It provides specialized installation and usage documentation, allowing developers to seamlessly integrate it into their familiar toolchains.

Practical Application Value: Empowering Efficient Collaboration for Individuals and Teams

• Individual developers: Structurally manage project ideas to avoid getting lost during development;
• Team collaboration: Establish a unified specification language to promote smooth communication between humans and AI agents;
• Lower barriers: Encode best practices into skills, enabling inexperienced developers to follow industry-standard processes.

Technical Implementation Features: Directive Document-Driven Methodology

The open-canal codebase consists entirely of directive documents with no execution scripts. Detailed process rules are stored in the standards directory; SKILL.md serves as a skill trigger and operation guide; the references directory stores long-form background materials. Skill directories follow the Agent Skills specification naming to ensure compatibility with the skill registry.

Summary and Outlook: Exploration of Normative Paradigms for AI-Assisted Development

open-canal solves the problems of context loss and planning chaos for AI coding agents in complex projects, representing an important exploration direction in the field of AI-assisted development—not only using AI to write code but also using AI to manage project specifications. In the future, such specification methodologies will become key support for AI agents with enhanced capabilities, driving a new development paradigm where humans and AI collaborate through a shared specification language.