
create-issues
by silver2dream
AI-assisted development workflow kit for Spec → Implement → PR → Merge automation
SKILL.md
name: create-issues description: Systematic GitHub issue creation from requirements. Analyzes codebase, breaks down requirements into multiple issues with dependencies, proposes detailed issues for review, and creates them after user approval. Triggers: /create-issues, create issues, batch issue creation, requirement breakdown. allowed-tools: Read, Grep, Glob, Bash
Create Issues Skill
Systematic workflow for creating well-structured GitHub issues from high-level requirements.
Overview
This skill transforms user requirements into a set of well-structured, dependency-aware GitHub issues. It follows a 5-phase workflow with a mandatory user approval gate before creating any issues.
When to Use
Use this skill when:
- User invokes
/create-issues - User describes a feature or requirement that needs multiple issues
- User wants to plan work breakdown systematically
- User needs help decomposing a complex requirement
Workflow
Phase 1: Codebase Analysis
Read phases/analyze.md
Scan the project to understand:
- Repository structure and languages
- Module organization and patterns
- Existing test coverage
- Configuration from workflow.yaml
Phase 2: Requirement Breakdown
Read phases/breakdown.md
Systematically decompose the requirement:
- Identify discrete tasks
- Detect dependencies between tasks
- Assign complexity scores (1-5)
- Assign priorities (P0/P1/P2)
- Group by execution phase
Phase 3: Issue Proposal Generation
Read phases/propose.md
Generate detailed proposals:
- Full issue content with all fields
- Scope with specific file paths
- Acceptance criteria
- Verification commands
- Dependency graph visualization
Phase 4: User Approval Gate
Read phases/approval.md
Present proposals and wait for confirmation:
- Display formatted proposals
- User responds: yes/no/edit
- Only proceed if user explicitly approves
Phase 5: Issue Creation
Read phases/create.md
Create issues via GitHub CLI:
- Execute
gh issue createfor each issue - Capture returned issue numbers
- Update dependency references
- Generate final report with links
Critical Rules
- NEVER create issues without user approval
- ALWAYS show full proposal details before asking for approval
- Respect dependency order - issues with dependencies must reference actual issue numbers
- Follow commit format -
[type] subject(no colon) - Include verification commands - Go projects use
go test ./...andgo build ./...
Integration with AWK Workflow
Issues created by this skill will have the ai-task label and can be processed by:
awkit kickoff- AWK principal workflow (spec-driven)/run-issues- Batch issue processing skill
⚠️ WARNING: Do NOT run /run-issues and awkit kickoff simultaneously on the same issues.
Recommended workflow:
/create-issues → (user approves) → issues created
↓
Choose ONE:
├─ awkit kickoff (structured, spec-driven)
└─ /run-issues (autonomous, batch processing)
Self-Check
On each phase entry, output:
[CREATE-ISSUES] <timestamp> | <phase> | loaded: <filename>
Quick Reference
| Phase | Action | File |
|---|---|---|
| 1. Analyze | Scan codebase structure | phases/analyze.md |
| 2. Breakdown | Decompose requirement | phases/breakdown.md |
| 3. Propose | Generate detailed proposals | phases/propose.md |
| 4. Approval | Wait for user confirmation | phases/approval.md |
| 5. Create | Execute gh issue create | phases/create.md |
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
1ヶ月以内に更新
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon

