yaml-kanban

SKILL.md

name	yaml-kanban
description	Use when editing kanban/TODO.yaml or kanban/DONE.yaml

YAML Kanban

Trigger skill for any kanban/TODO.yaml or kanban/DONE.yaml. Never write files for this skill other than kanban/TODO.yaml and kanban/DONE.yaml.

Schema

• Each file is a map from stable task id to task object.
• Never use ids outside t_<uint> shape, e.g. t_4; for new tasks, use the highest numeric suffix across both files plus 1.
• Never store task state in fields; infer it from filename: TODO.yaml is open, DONE.yaml is finished.
• Never change ids when moving or renaming tasks.
• Never add checkbox/status fields unless user asks.
• Never require start:; omit it until work actually starts.
• Never set start: or start agent e2e work without first reading human-written success_criteria: in TODO.yaml; demand it if absent.
• Never omit timezone from timestamps; prefer ISO, e.g. 2026-04-30T15:38:39+0300.
• Never duplicate completed work; use fixed_in: to point at the task id that fixed it.
• Never use prose dependency notes; use deps: with task ids.
• Never invent extra fields; preserve existing extra fields after deps:.
• Never reorder task fields; use title, success_criteria, start, finish, fixed_in, deps, then preserved extra fields.

t_1:
  title: fix responsible hono
  success_criteria: fix is verified by regression test
  start: 2026-04-30T12:02:00+0300
  finish: 2026-04-30T15:38:39+0300
  fixed_in: t_4
  deps:
    - t_2

Operations

• Add: create t_<next> in TODO.yaml with title:.
• Start: set start:.
• Finish: move the whole task to DONE.yaml and add finish:.
• Reopen: move the whole task to TODO.yaml and remove finish: unless user wants history retained.
• Rename: change only title:.
• Add dependency: append the dependency task id under deps:.
• Prioritize: read kanban/TODO.yaml and kanban/DONE.yaml, then reorder task keys in TODO.yaml by explicit deps: or inferred repository dependencies.
• Never skip YAML validation when feasible.