walk-me-through

Present verified research findings conversationally one claim at a time.

Updated Apr 11, 2026
One-click install
npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill walk-me-through-gozonerd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: walk-me-through
Source: https://github.com/gozonerd/shadow-ai-assessment/tree/main/.claude/skills/walk-me-through
Command: npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill walk-me-through-gozonerd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Complex research findings, assessment results, and data reports often arrive as dense walls of text that are hard to process, especially for users with cognitive load constraints like ADHD or migraine sensitivity. This Skill converts that material into a verified, bite-sized conversational walkthrough. ## Core Features & Use Cases - Claim Extraction and Deduplication: Reads every source in full, extracts discrete claims, merges them, and removes duplicates while flagging numeric discrepancies. - Multi-Pass Audit Verification: Runs repeated audit loops against the original sources until 5 consecutive null-edit passes confirm the claims list is accurate, catching hallucinations, misattributions, and omissions. - Conversational Walkthrough: Presents one finding per message in plain language, waits for the user's reaction, adapts sequencing to their interests, and captures their real-time insights. - Use Case: After receiving an 8-assessor evaluation report, ask for a walkthrough and receive each verified finding individually, with disagreements between assessors flagged explicitly rather than smoothed over. ## Quick Start Walk me through the findings from the attached research report, one finding at a time.

Frequently Asked Questions about walk-me-through

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

FAQPage Schema
How do I get research findings explained one at a time?

Ask the assistant to walk you through the findings conversationally. The skill extracts every claim from your sources, verifies them through repeated audit passes, then presents one finding per message, waiting for your reaction before continuing.

How does the walkthrough verify findings are accurate?

It runs an audit loop comparing every extracted claim against the original sources, checking for factual errors, misattributions, omissions, and duplicates. Presentation only begins after 5 consecutive full audit passes produce zero edits.

Can I use this for explaining general concepts or tools?

No. This skill is scoped to walking through research findings, assessment results, and data the user has not yet processed. Explaining standalone concepts or tools should use standard explanation patterns instead.

What happens when sources disagree with each other?

Disagreements are flagged explicitly rather than smoothed into false consensus. For example, it reports that 6 of 8 assessors said STRONG and 2 said MODERATE, along with the reasons for the split.

Why does the walkthrough take time before presenting findings?

Phases 1 and 2 extract, deduplicate, and audit all claims before any presentation. This prevents hallucinated or duplicated claims from reaching you, so the upfront verification delay is intentional.