Preface

People building AI Agent products have likely encountered this scenario: a user comes with a question they haven’t fully thought through, and the model either gives a direct answer or simply asks a clarifying question and stops. The former deprives the user of the thinking process, while the latter fails to make progress. Alternatively, the model might offer advice, but the user actually just wanted to organize their thoughts, not have someone else make decisions for them.

The “Socratic Questioner” (苏格拉底提问器) on SkillHub, introduced below, is a pure prompt orchestration designed specifically for this need to “clarify thinking.” It does not answer, advise, or judge; it only prompts the user to surface their own reasoning chain through layered questioning.

What is this

In one sentence: A Socratic questioning Skill with zero external dependencies and pure prompt orchestration.

  • Publisher: SkillHub community publisher user_fa14b5bc, categorized under ai-agent, version 1.0.0, tagged as community source.
  • Problem Solved: Distinguishing between “helping users think clearly” and “making decisions for users.” It does not do the latter; it only does the former.

Core Features

1. Six-Dimensional Questioning Framework. The framework covers six angles: Definition, Premise, Evidence, Impact, Boundary, and Alternative. In each round, the Skill selects a question from the “weakest” angle in the user’s current response, rather than following a fixed sequence of the six dimensions.

2. Three Levels of Questioning Depth. It offers three tiers: Shallow (4 rounds), Standard (6 rounds), and Deep (unlimited rounds). The tier determines when the questioning stops.

3. Closing Rules. There is only one termination condition: the user says “I understand” (我懂了). Upon closing, the Skill performs only two tasks—organizing the entire questioning chain and listing the key insights the user derived during the conversation. It adds no new content.

4. Iron Rules. Only ask, never answer; no advice, no evaluation, no conclusions. This rule is written in the description and acts as a hard constraint for the Skill itself.

5. Dependency Status. requires_api_key=false, zero external dependencies; all capabilities derive from prompt orchestration.

Installation and Enablement

It should be noted that: From the retrieved official materials, specific installation commands, skills directory paths, or configuration fields were not observed, nor were there specific instructions for Cursor / Codex CLI / Claude Code. Therefore, unverified commands are not provided here. Please refer to the official page:

https://api.skillhub.cn/user_fa14b5bc/socratic-questioner

This Skill is based on the general SKILL.md format. The page tags show requires_api_key=false, meaning no additional API Key configuration is needed to enable it.

Typical Usage

Usage is the dialogue itself; no configuration writing or parameter passing is required. There are only entry parameters and process rules:

1. Entry Input. The user provides a question they haven’t fully thought through, such as “I’m considering whether to change jobs.” At this point, the Skill will not provide analysis but will select the weakest link from the six dimensions to initiate the first question—for example, confirming what “change” specifically refers to in the context of “changing jobs” (Definition dimension), or “under what circumstances would you feel you have thought it through” (Boundary dimension).

2. Process Progression. For every round the user answers, the Skill locates the weakest angle in that response and asks the next question. For instance, if the user says “because the current team atmosphere is bad,” the Skill might ask what evidence supports this judgment (Evidence dimension), rather than offering comfort or agreement.

3. End and Output. If the user says “I understand” at any round, the Skill stops questioning and outputs two parts: the entire questioning chain and the key insights the user arrived at during the dialogue. If the user never says “I understand,” the Shallow tier ends after 4 rounds, and the Standard tier after 6 rounds; the Deep tier has no round limit and is terminated by the user actively.

Applicable Scenarios & Notes

Suitable for: Users who need to structure vague ideas themselves—career choices, trade-offs between plans, or organizing themes before writing. It is also suitable as a component of an Agent to enforce a round of reasoning clarification before “action.”

Notes:

  • It does not produce answers. If the user wants actionable advice or conclusions, this Skill cannot help.
  • It does not perform external retrieval or connect to tools. All questioning is built upon user input; it will not introduce new facts.
  • The “weakest angle” is determined by the prompt and is the model’s own judgment. In the Deep tier, users need the ability to identify whether the questioning truly hits the mark; otherwise, they may fall into an invalid loop.
  • The score and download counts are data captured at the time of retrieval (Score 100000, Downloads 1171, Installations 0) and may differ from the real-time display on the page.

Conclusion

The value of the Socratic Questioner lies not in its many functions, but in its strong constraints—only asking and never answering, until the user reaches a conclusion themselves. For scenarios requiring “clarification before action,” it is a very low-cost puzzle piece.

Official Address: https://api.skillhub.cn/user_fa14b5bc/socratic-questioner