sw-agiledevelopment: An Agile Development Skill Framework for AI Programming Agents

sw-agiledevelopment is a structured skill framework designed to address the problem that AI programming agents lack standardized software development processes. It converts mature agile development methodologies into AI-executable steps through the SKILL.md workflow, covering the complete process of requirement clarification, technical specification, test-driven development (TDD), system debugging, etc. It helps AI programming evolve from "writing code" to "doing engineering", improving the engineering quality of AI-generated code.

With the rapid development of AI programming assistants and agents, current tools often focus on code generation itself, ignoring key links in software engineering such as requirement analysis, architecture design, and test verification. The sw-agiledevelopment project proposes a solution to this problem, converting agile development processes into systematic steps that AI agents can follow.

Design Philosophy

The core of the framework is to combine mature software engineering practices with AI capabilities, using "skills" as the basic unit. Each skill corresponds to a specific link in the development process, with clear inputs, outputs, and trigger conditions, enabling AI agents to complete development tasks step by step instead of replacing human developers.

Multi-Platform Support

Natively supports OpenCode (plugin mode), Codex (plugin mode), GitHub Copilot (skill mode), and Android Studio AI Agent. Each platform has corresponding installation guides (e.g., OpenCode requires following .opencode/INSTALL.md).

Workflow

From requirement clarification to branch cleanup, it covers stages such as: requirement clarification → technical specification → work plan → development (including TDD) → code review → task verification → branch closure.

Core Skills

It includes 11 core skills, such as:

• sw-requirements-clarification: Convert ideas into business requirements when starting a new feature;
• sw-test-driven-dev: Enforce TDD cycles;
• sw-code-review: Conduct self-review after task completion;
• sw-systematic-debugging: Structured handling of bugs, etc. Each skill has clear purposes and trigger conditions.

TDD Practice

The sw-test-driven-dev skill enforces the red-green-refactor cycle: first write a failing test (red) → code to make the test pass (green) → refactor the code (keep the test passing), ensuring code testability and design quality.

Code Review

The sw-code-review skill simulates human review processes, checking code style, logic, security, etc., and outputs structured feedback.

Systematic Debugging

sw-systematic-debugging guides AI to adopt a structured method (understand the problem → hypothesize → verify → analyze) to avoid blind modifications; sw-parallel-debugging supports parallel processing of multiple independent failures.

sw-agiledevelopment is open-sourced under the MIT license, inspired by the Superpowers skill format and software engineering practices. The future roadmap includes integrating visual requirement analysis, enhancing analysis and documentation capabilities in the requirement phase, and moving towards more comprehensive software engineering support.

sw-agiledevelopment provides a standardized agile development framework for AI programming agents, helping to improve AI programming quality. As AI programming capabilities improve, such frameworks will become key support for AI to evolve from "code generators" to "engineering assistants", and are worth trying for teams.