AgentSkillsCN

what-to-work

通过 DeepWiki 与 Context7,对 BD 任务进行分类与梳理,明确下一步的工作方向。当用户希望深入了解、学习某个库、软件包或框架,或快速掌握其核心功能时,可使用此方法——例如:“请解释一下 React Query 是如何工作的”、“帮我理解 Zod 的原理”、“这个库到底有什么作用?”

SKILL.md
--- frontmatter
name: what-to-work
description: Decide what to work on next by triaging bd tasks, bundling into up to three options, or routing to plan-new-task when backlog is empty or too small

What To Work

Goal

Provide clear, user-confirmable next work options. If the backlog is too small to form a meaningful bundle, proactively route to plan-new-task (which uses new-task) to create new work.

Sizing rubric (expected code change, added + deleted)

  • tiny: ~10 lines
  • medium: ~100 lines
  • large: ~250 lines
  • xlarge: ~500 lines

Workflow

  1. Scan backlog (do-work guidance)

    • Run bd ready in all relevant project roots.
    • If multiple known repos exist, follow do-work guidance to scan them all.
    • Use bd show on top items to estimate size and dependencies.
    • Capture id, title, priority, type, dependencies, and size estimate (tiny/medium/large/xlarge).
  2. Decide if backlog is bundle-worthy

    • Bundle-worthy means you can form at least one large or xlarge bundle.
    • Target xlarge (~500 lines) and do not exceed xlarge.
    • If only tiny/medium or admin/event tasks remain, or total expected change is below large, treat the backlog as too small.
  3. Route

    • If bundle-worthy, use what-we-have-to-work to propose 1-3 bundles.
    • If empty or too small, use plan-new-task to propose 2-4 new feature ideas based on bd list and user context.
  4. Present next action

    • State which route you are using and why.
    • Ask for any missing context only if it blocks routing.

Output Template

Route: what-we-have-to-work

text
Backlog check complete: [summary]
Route: what-we-have-to-work
Next step: I'll bundle tasks into up to three options for you to pick.

Route: plan-new-task

text
Backlog check complete: [summary]
Route: plan-new-task
Next step: I'll propose new features based on bd list and your context.

Integration

  • Use do-work Phase A guidance for scanning repos and sizing bundles.
  • Use what-we-have-to-work when you can form a large/xlarge bundle (cap at xlarge).
  • Use plan-new-task when backlog is empty or too small; plan-new-task uses new-task for decomposition.
  • After user selection, follow do-work for execution.