mdtask has been introduced — a new CLI tool and Spec-Driven Development (SDD) methodology that allows managing tasks and technical documentation directly within a Git repository using Markdown files.
What Happened
mdtask implements an approach where all tasks are stored as checkboxes within Markdown files. This allows for synchronizing code changes, specification updates, and task closures within a single atomic commit. For AI agents, a 'skills' system is provided, allowing them to work cyclically: selecting a task, implementing it, updating documentation, and committing the changes.
Context
The tool implements the Spec-Driven Development (SDD) paradigm, turning the repository into a single source of truth. This solves the problem of "stale docs," where code changes while process descriptions become outdated, and minimizes the need for AI agents to interact with external task tracker APIs via MCP by replacing them with local context reading.
Why It Matters for the Industry
For the development industry, this signifies a shift toward Agent-native development, where project state management becomes native to Git. This increases the scalability of autonomous agents and standardizes a process where the task and its description are an integral part of the code rather than an external artifact.
Why It Matters for Users
Developers gain the ability to manage their backlog directly within their code editor without switching between Jira, Linear, and their IDE. For those using AI agents (such as Claude or Cursor), the process becomes simpler and more cost-effective, as it is easier for agents to read and update local files than to work with external services.
What Is Not Yet Known / Limitations
Questions remain regarding the tool's scalability in large Enterprise structures, where implementing such methodologies may require solving complex organizational challenges.
Sources
Author
Look at AI, Editorial Team