| name | logvalet:spec-to-issues |
| description | Decompose a specification document into Backlog issues: the LLM analyzes a spec/requirements file, breaks it into appropriately-scoped tasks, and creates them in Backlog one by one with dry-run preview before each submission. TRIGGER when: user says "specから課題作成", "仕様から課題を作って", "spec to issues", "仕様書を課題に分解", "要件を課題にして", "spec 分解", "仕様分解", "仕様書からチケットを切って", "spec からタスクを作成", "要件定義を課題に", "仕様をバックログに登録", "spec を issue に変換", "設計書から課題を起こして", "PRD から課題を作成", "タスク分解", "break down spec into issues", "create issues from requirements", "一括課題作成", "bulk issue creation from spec". DO NOT TRIGGER when: user wants to create a single issue manually (use issue-create) or wants to write a spec first (use an external spec tool).
|
logvalet-spec-to-issues
spec ファイルを解析し、適切な粒度の課題リストに分解してから Backlog に順次登録する。
For full logvalet CLI documentation, see the logvalet skill.
When to use this skill
Use logvalet-spec-to-issues when you need to:
- break down a feature spec or requirements document into actionable Backlog issues
- bulk-create a structured set of tasks from a design document
- convert a markdown spec file into a hierarchical epic → task structure
- onboard a new feature by registering all required tasks at once
Workflow
Step 1: Obtain spec content
Ask the user for one of:
- a file path (e.g.,
./docs/spec.md, ./requirements.md)
- pasted spec text directly in the conversation
Read the spec file if a path is provided.
Step 2: Confirm project and defaults
Ask the user (in a single question):
project-key — target Backlog project
- default
assignee (optional — can be left unset for all issues)
- default
priority (optional — default: "中")
- default
issue-type (optional — default: "タスク")
- whether to use parent/child structure (epic → task) if the spec has clear phases or sections
If project key is unknown, list available projects:
lv project list -f md
Step 3: Decompose spec into issue list
Analyze the spec and generate a proposed issue list. Guidelines:
- Granularity: Each issue should represent 1–3 days of work
- Title: Clear, actionable, specific (not "Implement feature X" but "X の API エンドポイントを実装")
- Description: Reference the relevant section of the spec (e.g., "spec §3 の認証フロー")
- Priority: Infer from criticality, dependencies, and user-stated defaults
- Dependencies: Note which issues depend on others
- Parent/child: Group related sub-tasks under a parent issue when a section is large
Step 4: Present issue list for review
Show the proposed issue list before creating anything:
## Spec → Issues 分解結果 — PROJECT_KEY
> spec: ./spec.md | 課題数: N件
### 課題リスト
1. **[高] API エンドポイント設計**
- 説明: spec §3 の認証エンドポイントを実装
- 推定工数: 2日
- 依存: なし
2. **[中] データベーススキーマ設計**
- 説明: spec §4 のテーブル定義に従いマイグレーション作成
- 推定工数: 1日
- 依存: なし
3. **[高] フロントエンド認証フロー**
- 説明: spec §5 のログイン/ログアウト UI
- 推定工数: 3日
- 依存: 課題1
---
N件の課題を作成します。[全て作成/選択/キャンセル]
Allow the user to:
- approve all
- select specific issues to create
- edit titles, priorities, or descriptions before creation
- cancel
Step 5: Dry-run preview
For each issue to be created, run --dry-run first:
lv issue create \
--project-key PROJ \
--summary "API エンドポイント設計" \
--description "spec §3 の認証エンドポイントを実装" \
--priority "高" \
--issue-type "タスク" \
--dry-run
Show the dry-run summary to the user and confirm before proceeding to actual creation.
Step 6: Create issues sequentially
Create each confirmed issue one by one:
lv issue create \
--project-key PROJ \
--summary "API エンドポイント設計" \
--description "spec §3 の認証エンドポイントを実装" \
--priority "高" \
--issue-type "タスク"
For child issues, use --parent-issue-id after the parent is created:
lv issue create \
--project-key PROJ \
--summary "子タスクのタイトル" \
--parent-issue-id PARENT_ISSUE_ID
Step 7: Report results
After all issues are created, show a summary:
## 作成完了 — N件の課題を登録しました
| # | 課題キー | タイトル |
|---|----------|----------|
| 1 | PROJ-101 | API エンドポイント設計 |
| 2 | PROJ-102 | データベーススキーマ設計 |
| 3 | PROJ-103 | フロントエンド認証フロー |
Notes
- This skill is SKILL-only — there is no dedicated
logvalet command for spec decomposition; the LLM does the analysis
- Granularity matters: too coarse (1 issue for entire feature) and it's untrackable; too fine (1 issue per function) and it's noise
- Reference spec sections in descriptions so issues stay traceable to requirements
- Creation order: create parent issues before children so
--parent-issue-id is available
- If the spec is very long (>500 lines), consider processing it in sections and asking the user to confirm between sections
Anti-patterns
- Do not create issues without showing the list and getting user confirmation first
- Do not skip
--dry-run before bulk creation
- Do not create issues with generic titles like "実装する" — always be specific
- Do not create more than ~20 issues in a single session without checking with the user
- Do not guess the project key — always confirm