Task Execution Skill

Recognizes when the user wants to execute decomposed beads tasks and invokes the /execute command.

Detection

Trigger when the user expresses intent to execute planned/decomposed work:

Direct requests:

• "execute the tasks", "run the tasks", "dispatch the tasks"
• "start executing", "execute the plan", "run the plan"

After decomposition:

• "now implement these", "go ahead and build it"
• "start working on them", "execute all of these"

Backlog execution:

• "work through the ready tasks", "execute what's ready"
• "run the next batch", "continue execution"

Action

Invoke /execute with appropriate flags based on context:

• If user mentions a specific epic: add --epic {id}
• If user mentions a specific task: add --task {id}
• If user says "fast" or "no review": add --no-review
• If user says "no commits": add --no-commit
• If user says "all at once" or "no stops": add --auto

The /execute command automatically reads the project's framework setting from .claude/task-decomposer.local.md and adapts subagent prompts and review criteria accordingly. No framework flag is needed — it's resolved from persisted project settings.