claude-consult

Records side-by-side consults comparing Claude and Grok answers with verbatim quotes.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Second opinions collapse too easily: asking two models the same question and remembering only one answer loses the exact point where they disagree. This Skill produces a consult record that keeps both answers visible, verbatim, and anchored to their sources so the disagreement stays inspectable. ## Core Features & Use Cases - Frozen question capture: Records the exact question wording (or both wordings if it shifted) so neither answer is compared against a question it was never asked. - Verbatim answer anchoring: Keeps load-bearing passages from each model with source and date, labeling paraphrase wherever compression happens. - Agreement map: Separates agree-on, split-on, and only-one-says lines instead of blending both answers into a single summary. - Use Case: You asked Grok and Claude whether to migrate a service to a new database. Use this Skill to produce one markdown record with both answers quoted, the split points mapped, and the net takeaway for your decision. ## Quick Start Ask the assistant to record a consult comparing the Claude answer and the Grok answer you paste in, keeping both verbatim with an agreement map and net takeaway.

Frequently Asked Questions about claude-consult

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

FAQPage Schema
How do I compare answers from Claude and Grok on the same question?

Paste both answers with their sources and dates, then freeze the exact question wording. The consult record quotes load-bearing passages from each side, maps where they agree and split, and ends with a net takeaway for your decision.

What goes into a second opinion consult record?

A consult record contains the frozen question, both answers with verbatim passages and sources, an agreement map split into agree-on, split-on, and only-one-says lines, a net takeaway, and a list of what could not be verified.

Can I record a consult with only one model's answer?

No. A record with only one answer is an answer note, not a consult. The Skill requires both answers present with sources; a missing or unreachable answer produces an incomplete record naming what is missing, never a verdict from one side.

What if the question wording changed between the two asks?

Record both wordings and treat the difference as a finding. Comparing answers to differently worded questions can make the split an artifact, so the wording shift is flagged before any comparison.

Does the consult skill re-ask the models or judge which is better?

No. It is strictly read-only: it never re-prompts a model, edits an answer, or rates which model is better in general. It only records and aligns answers that were already given.