Show HN: VS Code Agent Kanban: Task Management for the AI-Assisted Developer

Software engineer Gareth Brown has released VS Code Agent Kanban, a new extension designed to solve the persistent problem of 'context rot' for developers using AI coding assistants like GitHub Copilot. The tool addresses the common frustration where valuable planning conversations, trade-off analyses, and implementation decisions are lost when a chat session ends or hits context limits. Instead of relying on volatile chat histories, Agent Kanban creates a permanent, editable record by storing each task as a Markdown file within a dedicated .agentkanban/tasks/ directory.

Each task file features YAML frontmatter to track metadata like title, Kanban lane, and timestamps, while the body contains a structured log of the conversation between user and agent, marked with [user] and [agent] tags. This approach provides a readable, searchable source of truth that developers can commit to Git, offering a full history of the 'why' behind code changes. The extension integrates with existing agent harnesses through @kanban commands—'plan', 'todo', and 'implement'—to structure workflows without locking users into a proprietary system. By leveraging simple text files, it ensures tasks are diffable, mergeable, and accessible to entire teams, finally bringing persistent memory and coordination to AI-assisted development.