research-desk

Answers one hard question with three quoted sources and a falsifier.

Updated Sep 11, 2026
One-click install
npx skills add https://github.com/JYeswak/grok_bot_playground --skill research-desk-jyeswak
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: research-desk
Source: https://github.com/JYeswak/grok_bot_playground/tree/main/plugin/skills/research-desk
Command: npx skills add https://github.com/JYeswak/grok_bot_playground --skill research-desk-jyeswak

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When you ask a hard technical question, you often get vague answers, invented citations, or restated questions instead of verified facts. This Skill enforces a strict research protocol: one sentence answer, three quoted sources actually fetched from vendor documentation, and an explicit falsifier so you know what would change the answer. ## Core Features & Use Cases - Source-grounded answers: Fetches vendor documentation (preferring docs.x.ai/grok-bot/*) and quotes lines actually read, never invented. - Falsifiability check: Every answer ends with the one condition that, if false, would change the conclusion. - Honest failure mode: If the answer cannot be found, it returns "not found" plus every URL opened, instead of hallucinating. - Use Case: Ask "Does the Grok Bot API support streaming responses?" and receive a one-sentence answer, three quoted lines from fetched docs, and the falsifier that would overturn it. ## Quick Start Ask the research-desk skill a hard question such as "What rate limits apply to the Grok Bot API?" and it will return a sourced one-sentence answer with quotes and a falsifier.

Frequently Asked Questions about research-desk

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get a sourced answer to a technical question?▼

Ask the question in your own words and the skill fetches the relevant vendor documentation page, then returns one sentence answering it plus three quoted lines from pages actually opened. It never restates the question or invents citations.

What happens when the research skill cannot find an answer?▼

It returns "not found" along with a list of every URL it opened during the search. It never fabricates a quote or guesses an answer when sources are unavailable.

Does the research-desk skill require network access?▼

Yes, it needs network access to vendor documentation such as docs.x.ai, using an Accept header of */* since a text/markdown Accept header returns 404. No connectors or credentials are required.

Can the research-desk skill send emails or create routines?▼

No, it is strictly read-and-report. It never sends mail, creates routines, writes credentials, or publishes anything, so it is safe to run without approval workflows.

What counts as a failed research answer?▼

A greeting, a token echo, a restated question, or any source that was not actually fetched counts as failure. A passing answer is one sentence plus three verifiable quotes and a falsifier, or an honest "not found" with the URL list.