What is tbd?

tbd is a command-line tool designed to make AI coding agents more organized and consistent. It combines task tracking, planning templates, built-in engineering guidelines, and workflow shortcuts into a single tool that an agent can install and use across coding sessions. The task tracking part is based on a concept called beads, which are individual work items (bugs, features, epics) stored as Markdown files in your git repository on a separate dedicated branch. Because they live in git, they persist between separate agent sessions and across team members. The format avoids the merge conflicts and file locking issues that can arise with database-backed issue trackers. An agent can create, update, and query beads through simple CLI commands, and the tool installs itself as a skill in Claude Code so agents find it automatically. The planning workflow encourages writing a spec document before writing code, then breaking that spec into beads for each piece of work. tbd provides templates for creating plan specs, research briefs, and architecture documents. A shortcut system lets you ask an agent in natural language to do things like "create a PR" or "review this code," and the agent translates that into the appropriate tbd commands behind the scenes. The knowledge injection feature ships with over 20 guideline documents covering topics like TypeScript and Python best practices, test-driven development, monorepo setup, error handling, and backward compatibility. When an agent needs context on one of these topics, it loads the relevant guideline document rather than relying on its training data alone. tbd is aimed at developers who use AI agents heavily for coding and want more discipline around planning and quality. It works with Claude Code, Cursor, Codex, and other agent environments that can run a CLI tool. The full README is longer than what was shown.