Agent skill
dig
Clarify ambiguous points through structured questioning. Use when requirements are unclear, context needs investigation, or decisions need to be made. Triggers on "dig into", "clarify requirements", "investigate options".
Install this agent skill to your Project
npx add-skill https://github.com/yaakaito/env/tree/main/dotfiles/.claude/skills/dig
SKILL.md
Dig
Clarify ambiguous points in the current context through structured questioning.
Core Behavior
- Analyze - Read existing context (conversation, codebase)
- Identify gaps - Find ambiguous or incomplete points
- Ask questions - Use AskUserQuestion tool to clarify each item
- Iterate - Continue until requirements are clear
- Document - Update summary after each answer
Phase 1: Initial Analysis (Silent)
- Review conversation history
- Analyze relevant codebase sections
- Create a scope list of items needing clarification
- Categorize gaps across 8 areas:
- Scope, Behavior, Data, Users, Integration, Constraints, Priority, Edge cases
Phase 2: Iterative Questioning
MUST use AskUserQuestion tool (conversational questions are not allowed)
- 2-4 questions per round
- Each question has 2-4 options with pros/cons in descriptions
- Skip questions answerable from codebase analysis
- Do NOT guess or assume - always ask when unclear
Phase 3: Summary
After each round, update and display:
## Summary
### Confirmed
- [Item]: [Decision] - [Rationale]
### Pending
- [Item]: [Why it needs clarification]
Phase 4: Completeness Check
- Review all 8 categories for remaining gaps
- If gaps remain → Return to Phase 2
- If complete → Proceed to output
Output
Decisions Made
| Category | Decision | Rationale |
|---|---|---|
| ... | ... | ... |
Requirements
Functional:
- Requirement + acceptance criteria
Non-Functional:
- Requirement + metrics
Open Questions
- Items that still need resolution (if any)
Next Steps
- Plan mode - Break down into implementation tasks
Important
- Do NOT guess - Always ask when something is unclear
- Be thorough - Cover all relevant categories
- Stay focused - Only ask about the current context
- Track progress - Update summary after each round
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
spec-file-manager
Manage specification files. Use when creating, reading, updating, or listing spec files. Triggers on requests like "create spec", "update spec", "list specs", or "show spec status".
adr-writer
Skill for creating and editing Architecture Decision Records (ADRs). Use when requests involve "create ADR", "record architecture decision", or "document design decisions". Supports creating new ADRs, updating existing ADRs, and reviewing ADR format.
github-workflow
Write and edit GitHub Actions workflow files. Use when creating new workflows, editing existing .github/workflows/*.yml files, or setting up CI/CD pipelines. Triggers on requests like "create a workflow", "add GitHub Actions", "set up CI", or "edit the workflow file".
github-pr-unresolved-review-fetcher
Fetch unresolved PR review comments from GitHub. Use when working with pull request reviews, addressing review feedback, or checking outstanding review comments on a PR. Triggers on requests like "get PR reviews", "fetch review comments", "show unresolved comments", or "what feedback is pending on this PR".
github-workflow-fixer
Analyze and fix GitHub Actions workflow failures. Use when workflows fail, CI/CD issues occur, or when debugging GitHub Actions runs. Triggers on requests like "fix workflow", "why did the build fail", "check CI status", "debug GitHub Actions", or "workflow failed".
codex-review
Codex CLI を使って現在のコード変更を5つの観点(セキュリティ・パフォーマンス・バグ・テスト・アーキテクチャ)から並列レビューし、指摘に基づいてコードを修正するスキル。ユーザーが「codex でレビュー」「codex review」「コードレビューして」「レビュー実行」「codex にレビューさせて」「変更をチェック」などと言った場合や、コード品質チェック・多角的なレビューを求められた場合に使用する。明示的に codex に言及していなくても、包括的なコードレビューを求められたらこのスキルの使用を検討すること。
Didn't find tool you were looking for?