← Back to Skills Marketplace
inxan3

Interrogate

by inxan3 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
73
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install interrogate
Description
Guided elicitation skill. Activates ONLY when the user explicitly types /interrogate. Asks 4–15 adaptive questions in batches of 2, each with exactly 4 lette...
README (SKILL.md)

Interrogate

Activated by: /interrogate (literal command only — do not trigger on vague requests unless user types it)

Flow

1. Detect complexity

On /interrogate, read the rest of the message (if any) to understand the topic. If no topic is given, ask: "What are we figuring out?" before starting questions.

2. Ask in batches of 2

  • Present 2 questions at a time, each with options A / B / C / D
  • Always add a silent 5th option: user can ignore the letters and write freely — handle that naturally
  • After each batch of answers, decide:
    • Is the picture clear enough? → go to step 3
    • Need more clarity? → ask the next batch (max ~7 batches / ~15 questions total)
  • Keep questions tight and mutually exclusive — no overlap between options

3. Confirm before acting

When enough is understood, summarize in 3–5 bullet points:

  • What the user wants
  • Key constraints or preferences gathered
  • Anything still ambiguous (flag it)

Then ask: "Does this match what you have in mind? Say yes to proceed, or correct anything."

Only act after confirmation.

Question Design Rules

  • Options should be meaningfully different — not just degrees of the same thing
  • One option can always be "Something else entirely" if the space is truly open
  • Avoid leading questions — present options neutrally
  • Adapt next batch based on previous answers (tree-style branching, not fixed sequence)
  • For creative tasks: first batch should explore tone/style/audience
  • For technical tasks: first batch should explore scope/constraints/output format
  • For decisions/planning: first batch should explore goal/timeline/constraints

Format

**Q1. [Question]**
A) Option one
B) Option two
C) Option three
D) Option four

**Q2. [Question]**
A) Option one
B) Option two
C) Option three
D) Option four

*(Or just write what's on your mind — I'll adapt)*

Example opening batches

Creative task (/interrogate I want a landing page):

  • Q1: What's the primary goal of this page? A) Capture emails B) Explain a product C) Drive a purchase D) Build credibility
  • Q2: What's the tone? A) Clean and minimal B) Bold and energetic C) Warm and human D) Technical and precise

Vague task (/interrogate):

  • Ask "What are we figuring out?" first, then branch from the answer

When the user is undecided

If the user answers with "maybe", "not sure", "probably X or Y", or picks multiple options — do not re-ask. Instead:

  1. Briefly explain the key difference between the options they're torn between (1-2 sentences each)
  2. Add concrete trade-offs: cost, complexity, maintenance, limitations
  3. Give a clear recommendation based on context ("Given X, I'd go with B")
  4. Then let them confirm or override

Example: user says "A probably or maybe C" → "A (Railway URL) means it's always accessible from anywhere, zero effort, but uses a port on your Railway service. C (local/on-demand) means zero ongoing resource cost but you'd need to spin it up manually. Given you're already on Railway and want it always-on, I'd go with A."

Never leave the user hanging on a decision they don't have enough info to make.

Ending early

If the user gives a very detailed free-text answer at any point that makes further questions unnecessary, skip remaining questions and jump straight to the confirmation summary.

Usage Guidance
This skill appears coherent and low-risk: it only contains instructions for asking clarifying questions and asks for no credentials or installs. Before enabling widely, test it with a few harmless prompts to confirm the agent only activates on the literal `/interrogate` trigger and that it always summarizes and asks for confirmation before taking any action. If you rely on strict enforcement, consider limiting the agent's ability to autonomously invoke skills or monitoring logs to ensure it doesn't trigger in other contexts.
Capability Tags
cryptocan-make-purchases
Capability Assessment
Purpose & Capability
The name/description (guided elicitation activated by /interrogate) matches the SKILL.md behavior. No binaries, env vars, or config paths are requested — all proportional to a question/clarification helper.
Instruction Scope
The runtime instructions stay within the stated purpose (read the message, ask batches of 2 questions, summarize, require confirmation before acting). Two minor notes: the skill relies entirely on the agent obeying the literal trigger rule (only run on `/interrogate`), and the heuristics for 'picture clear enough' are left to agent discretion — which is reasonable but gives the agent operational choice about when to stop asking.
Install Mechanism
No install spec and no code files — lowest-risk instruction-only skill. Nothing will be downloaded or written to disk by the skill itself.
Credentials
No environment variables, credentials, or config paths requested. The skill does not ask for unrelated secrets or system access.
Persistence & Privilege
always:false (default) and no requests to modify other skills or system settings. disable-model-invocation is false by default (normal); this skill being user-invocable means it should only run when the user triggers `/interrogate`.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install interrogate
  3. After installation, invoke the skill by name or use /interrogate
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of the "interrogate" skill. - Guided questioning workflow: activated only by the explicit command `/interrogate`. - Presents 4–15 adaptive, mutually exclusive questions in batches of 2, each with options A–D plus a write-in option. - Adjusts questioning path based on prior answers, designed to clarify vague or complex requests before any action. - Summarizes understanding and asks for confirmation before proceeding. - Handles uncertainty and indecision by offering explanations and recommendations without repeating questions.
Metadata
Slug interrogate
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Interrogate?

Guided elicitation skill. Activates ONLY when the user explicitly types /interrogate. Asks 4–15 adaptive questions in batches of 2, each with exactly 4 lette... It is an AI Agent Skill for Claude Code / OpenClaw, with 73 downloads so far.

How do I install Interrogate?

Run "/install interrogate" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Interrogate free?

Yes, Interrogate is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Interrogate support?

Interrogate is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Interrogate?

It is built and maintained by inxan3 (@inxan3); the current version is v1.0.0.

💬 Comments