ShipSmooth: A Systematic Workflow Plugin for AI-Assisted Programming

ShipSmooth is a systematic workflow plugin for AI-assisted programming. Drawing on the spiral model and agile development principles, it provides AI assistants with a plan-driven, risk-prioritized, checkpoint-based agentic programming workflow. It aims to address problems in current AI-assisted programming caused by the lack of systematic guidance—such as unclear task priorities, insufficient risk identification, and difficulty tracking progress—by combining mature software engineering methodologies with AI agent technology to provide a structured collaboration framework.

As AI programming assistants become more capable, the way developers collaborate with AI is changing. However, current AI-assisted programming lacks systematic workflow guidance, which easily leads to chaos: unclear task priorities, insufficient risk identification, difficulty tracking progress, and hard recovery after session interruptions. The ShipSmooth project combines software engineering methodologies with AI agent technology to design a structured workflow framework to address these pain points.

The design of ShipSmooth is inspired by two classic methodologies:

1. Risk-driven thinking from the spiral model: Before developing each feature, identify and handle high-risk parts first, verify key technical difficulties early, and avoid discovering fundamental issues later.
2. Iterative delivery principle from agile development: Split tasks into vertical feature slices, each of which can be independently verified and delivered, improving efficiency and laying the foundation for continuous integration and delivery.

ShipSmooth defines a plan-driven execution process:

• Planning Phase: Create a Markdown-formatted plan file (including feature description, implementation strategy, task list) and submit it to version control to ensure traceability.
• Task Decomposition and Risk Assessment: Refine tasks into executable units, mark risk levels (high/medium/low), and developers can override AI assessments.
• Execution Phase: Execute tasks in descending order of risk. High-risk tasks use a two-stage process of de-risking/reinforcement (verify feasibility first, then optimize), while low-risk tasks are completed in one pass, with automatic Git commits created.
• Closure Phase: Mark the plan as completed, archive task status files, and perform a squash merge into the main branch to ensure a clean code repository.

ShipSmooth's key features include:

1. Risk Prioritization and Vertical Slicing: Risk management is prioritized; high-risk tasks are handled first to verify assumptions. Tasks follow the vertical slicing principle, where each task represents a complete user-perceivable feature.
2. Pause and Resume Mechanism: Plan and task status are persisted to the Git repository, supporting seamless pause and resume without re-establishing context.
3. Local Task Tracking: Uses pure local XML files to manage task status, synchronized with code version control, supporting offline use and data sovereignty, with optional integration with tools like Linear.
4. Progressive Code Quality Improvement: High-risk tasks initially focus on verifying ideas; after the solution is confirmed, refactoring and supplementary testing are done to avoid premature optimization.

ShipSmooth is provided as a plugin and supports multiple platforms:

• Claude Plugin: Register and install via the plugin market;
• Gemini Extension: Provides a dedicated extension installation method. When the plugin starts for the first time, it automatically installs dependencies and copies scripts to the local cache. Workflow specifications are defined in the SKILL.md file, providing execution guidelines for AI assistants.

ShipSmooth is not just a tool; it also demonstrates how to translate software engineering best practices into instructions that AI can understand and execute. As AI programming assistants become more capable, structured collaboration models will be key to improving efficiency and code quality. For organizations and developers looking to standardize team AI-assisted programming practices, ShipSmooth provides a well-thought-out starting point.