breakdown

Solid

Universal planning skill: Bug reads ROOT_CAUSE then estimates; Story/Task/Epic explores codebase then splits into sub-tasks with estimates, and packs each sub-task into a self-contained task.md work order for engineering to consume. Also handles scope challenge (advisory mode). Trigger: 拆單, 'split tasks', 拆解, 'breakdown', 'break down', 子單, 'sub-tasks', 評估這張單, 'evaluate this ticket', 估點, 'estimate', 'scope challenge', '挑戰需求', 'challenge scope', '需求質疑'.

AI & Automation 4 stars 0 forks Updated 2 days ago MIT

Install

View on GitHub

Quality Score: 80/100

Stars 20%
23
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
80
License 10%
100
Description 5%
100

Skill Content

# Breakdown — Packer `breakdown` 是 Packer:接收 refinement artifact、refinement Bug source mode RCA、JIRA ticket 或 DP source,把已定案的需求拆成可施工 work orders。它不擁有需求探索或技術決策;需要 改 Goal / Background / Decisions / Blind Spots / Technical Approach 時,route back to `refinement`。 ## Mandatory Contracts - 開始前讀 `workspace-config-reader.md`、`workspace-language-policy.md`、 `authoring-preflight.md` 與 root `language`;preview、JIRA comment、task.md / V*.md artifact 必須直接用 policy language 起稿,不可把 language gate 當送出前翻譯器。 - 寫 artifact 前必讀 `pipeline-handoff.md` § Artifact Schemas,再讀 `refinement-artifact.md` / `task-md-schema.md` 等對應 artifact-specific schema。breakdown 是唯一直接消費 `refinement.json` derive work order 的 owner;atom ownership 邊界以 `pipeline-handoff-atom-matrix.md` 為準,SKILL 主���不複製完整 schema 表。 - 寫入 specs Markdown 時遵守 `starlight-authoring-contract.md`;task work order 寫入 folder-native `tasks/Tn/index.md` 或 `tasks/Vn/index.md`;task schema 以 `task-md-schema.md` 為準。 - 所有 estimate 使用 `estimation-scale.md`;JIRA sub-task / story point 操作使用 `jira-subtask-creation.md` 與 `jira-story-points.md`。 - 寫入 task.md 前必須有 explicit user confirmation;沒有確認不可寫 JIRA、branch、 task.md、sidecar processed flag。 - DP-backed source 若由 `auto-pass` dispatch,explicit confirmation 可由 `AUTO_PASS_LEDGER_PATH=<absolute ledger path>` envelope token 提供,但 breakdown 必須先用 `scripts/validate-auto-pass-ledger.sh` 驗證 schema、source match、三個 consent boolean、 canonical `consent_excludes` enum 與 task write timestamp ordering。缺 token、r...

Details

Author
HsuanYuLee
Repository
HsuanYuLee/polaris
Created
4 months ago
Last Updated
2 days ago
Language
Shell
License
MIT

Integrates with

Similar Skills

Semantically similar based on skill content — not just same category

AI & Automation Solid

refinement

Iteratively enriches incomplete JIRA Epics into estimation-ready, technically-validated specs. Five modes: batch readiness scan, RD discovery (Phase 0), PM elaboration (Phase 1), technical approach (Phase 2), and multi-round iteration. Phase 1 goes beyond checklist filling — it explores the codebase, hardens AC, and produces a structured artifact for downstream skills. Trigger: "refinement", "grooming", "討論需求", "需求釐清", "補完 Epic", "這張單缺什麼", "brainstorm", "方案討論", "想重構", "tech debt", "batch refinement", "sprint prep", or Epic with sparse content needing enrichment.

4 Updated 2 days ago
HsuanYuLee
AI & Automation Solid

engineering

Engineer-minded execution orchestrator: takes a planned JIRA ticket and implements it with strict quality discipline — TDD, lint, typecheck, test, behavioral verify, PR. Two modes: first-cut (new implementation) and revision (fix PR review comments by returning to the work order). Local-only workflows may register delivery extensions, but those extensions are not part of the portable skill contract. Supports batch mode via parallel sub-agents. Trigger: "做 PROJ-123", "work on", "engineering", "開始做", "接這張", "做這張", "修 PROJ-123", "fix review on PROJ-123", PR URL (from pr-pickup or direct), or user provides JIRA ticket key(s). NOT for planning: Bug → refinement Bug source mode first; Story/Task/Epic → breakdown first. Key distinction: "下一步" / "繼續" without ticket key → my-triage (zero-input router + resume scan).

4 Updated 2 days ago
HsuanYuLee
API & Backend Listed

planning-and-task-breakdown

Task ledger: Use when ready decisions need one delegated, dependency-ordered tasks.md with owners, proof, cleanup, and reopen conditions. Own faithful ledger authoring; Skip when behavior remains unresolved, implementation is requested, or root planning-phase ownership and readiness review are needed.

6 Updated today
Dankosik