doughnut: Skill for Claude Code

Execute planned or planless work

Execute the selected plan or one explicitly authorized planless quick slice from an understood canonical feature story or contextual instruction. The coordinator owns delivery; implementation agents return uncommitted changes.

Establish execution context

Identify the execution source before changing project state:

• Planned: require an executable plan. For a feature story, read its seed section. For a bounded correction, require its complete correction input — its story and plan, or a plan-homed correction's plan — under planning scope and lifecycle.
• Quick: require an explicit current instruction that authorizes planless execution. The source is either a canonical feature story with understood goal, scope, key examples, and no blocking decision, plus the instruction to skip slice planning; or a contextual instruction that supplies the objective, known expectations, remaining uncertainty, and authority. A seed alone or apparent story size cannot select this path. Corrections require plans. In Story Branch or Trunk Mode, first admit an accepted independent mission; otherwise create no story, plan, or queue entry.

Name missing authority or source context and stop before backlog changes, delegation, or implementation. Successful quick execution keeps its scope in its story, when admitted, and its decisions, progress, and proof in the conversation; create no plan, completion note, or substitute record. A contextual instruction may leave expectations unresolved. Carry its goal and uncertainty, resolve what a behavior change requires, and allow an evidenced no-change conclusion through the explained-empty-change path.

Use planning level for source ownership, proof ownership when mapping or accepting proof, and active-plan refinement for plan updates. Judge proof during execution; retain completed plans, source history, and proof decisions for retrospective and story wrap-up. Quick work retains its source, conversation, and execution identity. Use slice decomposition for Behavior/Structure types and slice sizing for budgets and learning escalation.