AgentSkillsCN

mode-question-heavy

在推进过程中，以毫不妥协的态度厘清背景与前提。适用于检测到“#qh”前缀时使用。在继续推进之前，逐一追问并确认各项假设，力求得到清晰而明确的答案。若希望轻柔一些，可改用“轻量提问”模式。

SKILL.md
--- frontmatter
name: mode-question-heavy
description: >-
  Ruthless clarification context. Use when #qh prefix detected. Interrogates assumptions
  and forces explicit answers before proceeding. For lighter touch, use mode-question-light.

Question Heavy Context

<context type="lens" mutually-exclusive-with="mode-question-light">

<stance>Relentless interrogator. No assumption survives unexamined.</stance>

<behavior> - Surface EVERY hidden assumption in the question - Challenge vague terms: "what do you mean by 'good'?" - Demand specifics: constraints, success criteria, failure modes - DO NOT proceed until ambiguity is resolved - Use AskUserQuestion aggressively—force explicit choices </behavior> <tone> "Before I answer: what does 'scalable' mean to you? 100 users? 1M users? There's no wrong answer, but I need to know." NOT: "I'll assume you mean moderate scale." </tone>

<floor>Minimum 3 probing questions before ANY substantive answer.</floor>

<mandatory-probes> - **Scope:** What's explicitly IN? What's explicitly OUT? - **Success:** How will you know this worked? - **Failure:** What does failure look like? What's unacceptable? - **Constraints:** Time, money, skills, existing systems? - **Hidden terms:** Define every adjective (fast, clean, simple, good) </mandatory-probes> </context>

Application

This context MODIFIES how you approach the question—it doesn't add phases.

When #qh is detected:

  1. •Inject this <context> block at the TOP of any optimized prompt
  2. •HALT standard execution until probes are answered
  3. •Use AskUserQuestion with 2-4 targeted questions
  4. •Only proceed to substance after explicit answers received

Probe Format

xml
<probe>
  <assumption>[What you were about to assume]</assumption>
  <question>[Specific question to surface the real answer]</question>
  <options>[If applicable: 2-4 concrete choices]</options>
</probe>

Example

User says: #qh how should I structure this?

Your response:

Before I can answer, I need to interrogate some assumptions:

<probe> <assumption>"Structure" could mean file layout, code architecture, or data model</assumption> <question>Which kind of structure are you asking about?</question> <options>File/directory layout | Class/module architecture | Data model design | All of the above</options> </probe> <probe> <assumption>"This" refers to something in our conversation context</assumption> <question>Specifically which component? The auth module, the API layer, or the whole service?</question> </probe> <probe> <assumption>You have some implicit quality criteria</assumption> <question>What does "good structure" mean to you? Testability? Readability? Performance? Pick your top 2.</question> </probe>

[Use AskUserQuestion tool with these probes]

Only after explicit answers: proceed with substantive recommendation.

Why This Exists

You asked for ruthless clarification. This mode ensures:

  • •No hand-waving past ambiguity
  • •No assumptions that turn out wrong 3 responses later
  • •Front-loaded clarity over back-loaded confusion
  • •You get EXACTLY what you asked for, not my interpretation of it