From dev
Decompose prd.md into an ordered task list (tasks.yml). Mechanical phase — human reviews but agent drives. Triggers on: "dev/how", "decompose", "break this down", "create tasks"
How this skill is triggered — by the user, by Claude, or both
Slash command
/dev:howThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Consumes `.dev/prd.md`, produces `.dev/tasks.yml`. Interactive — the human reviews and adjusts the plan before it's finalised.
Consumes .dev/prd.md, produces .dev/tasks.yml. Interactive — the human reviews and adjusts the plan before it's finalised.
Read .dev/prd.md. If it doesn't exist, tell the user to run dev/what first.
The PRD's open questions section must be empty. If it's not, resolve them with the user before proceeding.
Break user stories into implementation tasks. Each task must be:
dev/build invocationRules of thumb — a task is too big if:
Right-sized examples:
Too big:
Dependencies first:
If task B reads from a table that task A creates, A comes first.
Things that must be true before Build starts (e.g. database running, env vars set). Each requirement gets a check command that can verify it.
Pull from the PRD's QA section. Split into:
agent: checks the build agent runs after each taskhuman: checks requiring human judgement (deferred to end)Walk through the full plan with the user:
Get explicit approval before saving.
Save to .dev/tasks.yml. See references/tasks-schema.md for the format.
Stage and commit .dev/tasks.yml so the tree is clean for dev/build:
chore: dev/how — [short feature name]
prd field must point to the PRD file useddev/what rather than guessing.dev/tasks.yml before finishingnpx claudepluginhub codethread/claude-code-plugins --plugin devGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.