dbs-good-question is a Chinese-language agent skill that improves how a question is asked. It pins a vague complaint such as “why does nobody buy” to something observable, finds the conflict inside it, and writes a problem brief with the object, goal, variables, constraints and feedback source an agent needs.
It also rates how far an agent can solve the task on four levels, from highly automatable to not yet suitable, and splits the work into what the agent can do, what needs human judgment and what needs real-world feedback.
What You Get
- The break point it sees in your question, in one or two sentences
- A clarity level of low, medium or high, and the single biggest gap
- A half-finished or full problem brief with the unknowns marked as unknown
- Up to two low-confidence candidate explanations with the signals that would confirm them
- An automation rating from A to D with a split of agent work, human judgment and feedback loops
- Up to three questions, only if the answers would change the judgment
Common Use Cases
- Reframing ‘why is nobody buying my course’ into a measurable drop between inquiries and payments
- Deciding whether an expense-approval process can be handed to an agent
- Preparing a clean problem statement before asking an AI to explain why saves are high but enquiries are low
What Makes It Unique
The skill treats the question as the thing to fix. It pins the phenomenon first, exposes the conflict, and gives candidate explanations only with a statement of what would confirm them. When information is short it writes a half-finished brief and marks the unknowns, leaving the gaps visible. It separates generating an explanation from solving the problem, since solving needs action and feedback. Scoring tables appear only when the user asks for an audit.
Under The Hood
It is a prose-only skill with no scripts. The agent reads the SKILL.md and loads two reference files, an output-format guide and a scenarios file, when a full template or a similar scenario is needed. It works from what you write or attach in the conversation. It does not run commands, call web services or write files.