skills.rest
elvisun
elvisunCOMMUNITY
Explore all skills by this author →
Website𝕏@elvissunGitHub

ai-visibility-panel-design

Selects QA-approved intent cells into a versioned AI-visibility tracking panel with weights and uncertainty rules.

663|47|Updated May 19, 2026
Search Tags:#panel design#weighting#ai visibility#stratified sampling#measurement plan#uncertainty intervals#campaign experiment
One-click install
npx skills add https://github.com/elvisun/newsjack --skill ai-visibility-panel-design
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ai-visibility-panel-design
Source: https://github.com/elvisun/newsjack/tree/main/skills/ai-visibility-panel-design
Command: npx skills add https://github.com/elvisun/newsjack --skill ai-visibility-panel-design

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Turning a pool of QA-approved prompt candidates into a defensible, statistically sound AI-visibility measurement plan is error-prone: teams mix lanes, invent weights, and overclaim attribution. This Skill converts accepted intent cells into a versioned tracking panel with explicit strata, weights, uncertainty methods, and refresh rules.

Core Features & Use Cases

  • Stratified Panel Selection: Allocates canonical intent cells across proximity bands, journeys, locales, evidence grades, and partitions (core, rotating, sentinel, control, aided) without peeking at baseline performance.
  • Honest Weighting & Uncertainty: Separates exposure and priority weights with provenance, and prescribes Wilson intervals, cluster bootstraps, and effective sample size reporting.
  • Versioning & Campaign Controls: Freezes panel versions with hashes and change ledgers, enforces treatment/control pre-registration before any causal campaign claims.
  • Use Case: After prompt QA approves 200 candidate prompts, use this Skill to select a 90-cell standard panel, assign evidence-backed weights, and emit tracking_plan.md, panel.yaml, and a run manifest template for Gate 4 human approval.

Quick Start

Use the ai-visibility-panel-design skill to build a versioned tracking panel from my QA-approved prompt candidates and measurement charter.

Frequently Asked Questions about ai-visibility-panel-design

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

FAQPage Schema
How do I design an AI visibility tracking panel from approved prompts?▼

Provide a measurement charter, prompt architecture, QA-approved candidates with a rejection ledger, and budget. The skill stratifies canonical intent cells across partitions and lanes, assigns evidence-backed weights, and outputs a tracking plan plus machine-readable panel.yaml.

What is the difference between core, rotating, sentinel, and control partitions?▼

Core is the continuously tracked set, rotating is a discovery set refreshed quarterly, sentinel acts as a tripwire for drift, and control is a false-positive check. Aided cells form a separate prompted partition never mixed with unaided denominators.

How are exposure and priority weights handled in panel design?▼

Exposure weights come from audience and intent prevalence evidence with source IDs; priority weights record human strategic decisions with approver artifacts. If credible exposure data is missing, equal weights within declared strata are used instead.

Can a before-and-after visibility increase prove campaign attribution?▼

No. A before/after increase alone is not attribution. The skill requires treatment/control definitions, pre-registration, and a credible experimental or counterfactual design before any causal language is allowed.

What uncertainty methods does the panel use for repeated prompt runs?▼

Wilson intervals apply only to simple unweighted strata with one independent observation per cell. With variants or repeats, it uses cell-cluster bootstrap or hierarchical methods, and stratified cluster bootstrap for weighted aggregates.

When should I not use percentage leaderboards for subgroups?▼

When a subgroup has fewer than 20-30 distinct cells, report raw counts and responses instead of percentages. Small denominators make percentage rankings misleading and unstable across waves.

Related Skills

Explore More Data & Analytics Skills→
slowman2084slowman2084
6

meta-reviewer

Evaluate meta-prompt-engineer outputs against ExpectedOutput fingerprints for cheating and overfitting.

Community
Advanced
bhanukushwahbhanukushwah

execute-phase

Execute development plans in wave-based parallel phases with sub-agents.

Community
Advanced
rocklambrosrocklambros
2

selecting-statistical-test

Select the appropriate statistical test from study design, measurement scale, sample size, and assumption checks.

Community
Advanced
HmbownHmbown

panel-grab

Capture a single panel to secure narrow lane control.

Community
Intermediate