deepgrain-foundry/.kandown/kandown_work.md

Deepgrain Foundry: working the board

The board is the owner's to-do list. docs/kandown.md explains how kandown works here.

• Work on tasks only when the owner asks ("do the next task", "do t8"). Don't pick up work on your own.
• Kinds of task.
  • Todo: things to build or fix.
  • Backlog, tagged discussion: give options with trade-offs and a recommendation, in chat, and write nothing in code until the owner decides. Then record the decision.
  • Backlog, tagged idea: ideas for later. Leave them alone unless asked.
  • Questions get answers.
• Keep the owner's words. Never rewrite or tidy text the owner wrote in a task file, and don't rename tasks the owner made. Add what you learn as ## Subtasks with report: lines, or in the report field.
• Writing tasks from chat. When you write a description, task or subtask from what the owner said in chat, write it fully in your own words as a clear description of what they want. Don't label it as the owner's, add dates, or quote them.
• When a task is finished, move it to Review (that is "done, please check") and set the report field: what changed, and how to check it in the game. Never move a task to Done: the owner does that after checking it. If the owner moves it back with feedback, keep the feedback and work from it.
• Log your work. Add an entry at the top of docs/agent-log.md with the date, the task id and title, what changed, the files touched, the tests run, and anything left open.
• Record decisions. When the owner decides something about the design, add it to the decisions list in docs/DESIGN.md with the date and the reason they gave.
• Commit task files with the work they describe. The owner may be editing tasks too: stage only the task files you changed.
• New tasks the owner asks for: kandown create "title" --to Todo (add -t discussion or -t idea for Backlog), then describe the task under ## Context.