dbs-good-question skill: rewrite a vague question into a clear brief and see how much an agent can solve
Rewrites a vague question into a clear problem brief and tells you how much of it an agent can solve.
- Installs
- 24.3K
- on skills.sh
- Stars
- 11K
- on GitHub
- Files
- 3
- besides SKILL.md
- Updated
- Oct 2026
- last repo commit
dbs-good-question is one skill from dontbesilent's dbskill toolbox. You give it a fuzzy question such as why a course does not sell, or a task you want to hand to an agent. It names the break in the story, rates how clear the question is (low, medium or high), and writes a draft brief with the object, goal, conflict, constraints and feedback source.
It works in four modes: a vague question, a situation with data, an automation check graded A to D, and candidate explanations. It asks at most 3 questions and ends with one small next step. SKILL.md is written in Chinese, and says to reply in the language you use.
Good for
- Founders and creators who ask big why-questions and get vague answers
- People deciding whether a task can be given to an AI agent
- Anyone who wants to turn a fuzzy problem into a brief another person or agent can work on
Install dbs-good-question
Claude Code
- 1
Copy this prompt.
Install the dbs-good-question skill in Claude Code. Run this command: claude plugin marketplace add dontbesilent2025/dbskill claude plugin install dbs-good-question@dontbesilent-skills Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Pick one route, not both. The README installs the whole 34-skill set with npx -y skills add dontbesilent2025/dbskill -g --all, which targets every agent the skills CLI finds; the routes here install only dbs-good-question. The README says plugin commands carry a prefix (for the full dbs plugin it is dbs:), so check the name Claude Code shows.
From the project README. Source.
What's in the skill
When it loads
把模糊提问改写成 Agent 可推理、可批评、可验证的问题说明书，并判断自动化可解程度。用户要改进问题表述、研究问题或评估 Agent 可解性时使用；任务交付目标已明确时不重复澄清目标。
Sections in SKILL.md
- 核心哲学
- 原则 1：好问题先钉现象
- 原则 2：好问题要暴露冲突
- 原则 3：Agent 需要约束场
- 原则 4：自动化解决需要反馈回流
- 原则 5：不要装确定
- 原则 6：先给抓手，再做审计
- 边界
- 工作模式
- 标准流程
- Phase 1：识别输入类型
- Phase 2：好问题五项检查
- Phase 3：判断 Agent 可解性
- Phase 4：改写成问题说明书
- Phase 5：生成候选解释并批评
- Phase 6：给一个最小下一步
Key facts
- Installs on skills.sh
- 24,296 (data from 2026-11-02)
- GitHub stars
- 11,039 for the whole dbskill repo (2026-11-02)
- License
- CC BY-NC 4.0 per README and LICENSE (non-commercial use only); the crawl lists no license identifier
- Last repo commit
- 2026-10-28
What it includes
Names the break, rates clarity, drafts a half-finished brief
Starts with one or two sentences on where the story breaks, then gives clarity (low, medium, high) and the biggest gap, before any scoring table.
Try: “为什么我的内容有人收藏但没人咨询？ (or in English: Why do people save my posts but never enquire?)”
Read ·
Mode A: vague questionFinds the core conflict and writes a full problem brief
Takes data, cases or chat logs, pulls out the conflict, and may add one or two low-confidence explanations with the signals you would see if they were true.
Read ·
Mode B: situation with dataGrades a task A to D for automation
A is highly automatable, B semi-automatable, C reasoning help only, D not yet suitable. It splits the task into what the agent can do, what needs your judgement and what needs feedback loops.
Try: “Can an agent automate our expense approval process?”
Read ·
Mode C: can an agent do it?Up to 3 explanations, each with a rebuttal and a test
Used when the question is already clear. Each explanation gets a mechanism, observable signals, what it rules out and what you would do differently.
Read ·
Mode D: candidate explanationsObject, goal, conflict, constraints, feedback, each scored 0 to 2
The full 10-point table is shown only when you ask for a strict audit or a score.
Read ·
Five-point question checkTemplates for audit, solvability, full brief and explanations
Read when you ask for an audit, an automation verdict or a full brief.
Read ·
references/output-formats.mdWorked examples for content, profile pages and automation
Shows how to handle cases such as saves without enquiries or an expense-approval automation request.
Read ·
references/scenarios.md
From the official docs
Questions
What does dbs-good-question do?
It turns a vague question into a problem brief with a clear object, goal, conflict, constraints and a way to check the answer. It also tells you how much of the problem an agent can solve.
How do I install it in Claude Code?
Use the install box at the top of this page. The dbskill README also installs all 34 skills at once with one skills CLI command, which targets every agent the CLI finds.
Does it only work in Chinese?
SKILL.md is in Chinese, but it says to answer in Chinese if you write Chinese and in English if you write English.
Does it give me the answer to my business problem?
No. It is built to clarify the question. SKILL.md lists confirmed cause diagnoses and solutions as outside its scope, and for emotional, health, legal or investment questions it names the limits and suggests a professional.
Does it need an API key, internet or scripts?
SKILL.md names none. It bundles no scripts.
Can I use it commercially?
The repo's LICENSE file is CC BY-NC 4.0, and its NonCommercial term reads: "You may not use the material for commercial purposes." If you are unsure whether your use counts as commercial, ask the author.
Is it safe?
It is plain instructions with no scripts. It reasons over what you paste into the chat, so leave out private data you do not want in the conversation.
Information
updated Nov 2, 2026
- Repository
- dontbesilent2025/dbskill
- Folder
skills/dbs-good-question
Sources
Docs, issue trackers and discussions this page is built from.
Show all 6Hide
- dbskill README · github.com
- skills CLI commands · github.com
- dbs-good-question SKILL.md · github.com
- dbs-good-question output-formats.md · github.com
- dbs-good-question scenarios.md · github.com
- dbskill LICENSE (CC BY-NC 4.0) · github.com