skills.rest
elchananteitzwork-a11y
elchananteitzwork-a11yCOMMUNITY
Explore all skills by this author →
GitHub

confidence-tiered-extraction

Enforces confidence-tier labeling and bans revenue, ROI, and success-probability claims in user-facing output.

Updated Jun 11, 2026
Search Tags:#scoring#claim validation#prompt compliance#evidence labeling#ai output review#confidence tiers#verdict formatting
One-click install
npx skills add https://github.com/elchananteitzwork-a11y/supplement-intelligence --skill confidence-tiered-extraction-elchananteitzwork-a11y
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: confidence-tiered-extraction
Source: https://github.com/elchananteitzwork-a11y/supplement-intelligence/tree/main/.claude/skills/confidence-tiered-extraction
Command: npx skills add https://github.com/elchananteitzwork-a11y/supplement-intelligence --skill confidence-tiered-extraction-elchananteitzwork-a11y

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It prevents prompts, extraction pipelines, and interpretation-formatting code from presenting unverified claims, revenue projections, ROI estimates, or success-probability language to founders, ensuring every user-facing statement carries an explicit evidence tier. ## Core Features & Use Cases - Confidence Tier Enforcement: Requires every claim, score, or verdict to be labeled CONFIRMED, INDICATED, or LIMITED inline, not hidden behind tooltips. - Forbidden Language Guardrails: Blocks TAM/SAM market sizing, revenue framing, ROI, break-even, payback, and success-probability phrasing per Constitution Laws 1-4. - Scoped Review Checklist: Applies only to prompt strings, extraction rules, and interpretation-formatting code in science-engine, regulatory-engine, ai-interpretation, prompts, consumer-intelligence, thesis-engine, verdict-matrix, and scoring paths. - Use Case: Before editing an Anthropic prompt that formats a product verdict for a founder, run this checklist to confirm each score carries a tier and no revenue or ROI language appears in the output. ## Quick Start Ask the AI to review the prompt or interpretation-formatting code you are editing against the confidence-tiered-extraction rules before committing it.

Frequently Asked Questions about confidence-tiered-extraction

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

FAQPage Schema
How do I add confidence tiers to AI-generated claims?▼

Label every claim, score, or verdict inline as CONFIRMED, INDICATED, or LIMITED based on evidence strength. The tier must appear directly in the output text, not behind a tooltip or info icon, and weak evidence must be stated plainly on the first screen.

What language is forbidden in founder-facing scoring output?▼

Revenue and market-size language (TAM/SAM, dollar market figures), success-probability phrasing (likely to succeed, Opportunity Score), and ROI terms (estimated ROI, break-even, payback period, margin projections) are all forbidden.

When should I apply the confidence-tiered-extraction checklist?▼

Apply it before writing or editing any Anthropic prompt string, extraction pipeline, or code that formats a verdict, score, or interpretation shown to a founder. Do not use it for provider fetch logic, unrelated UI copy, or test-only diffs.

Does this Skill review security or code quality issues?▼

No. It only covers content and code producing user-facing claims. Security review belongs to security-review, and code quality review belongs to code-review under independent-reviewer-agent.

Can this Skill delegate work to other agents?▼

No. It never invokes other Skills or spawns Agents, and its frontmatter disallows the Agent tool. If the work needs a different specialist, it stops and lets the Planner reassign the task.

Related Skills

Explore More Legal & Compliance Skills→
Norman-buryNorman-bury
3.0k

writing-core

Enforce de-AI-ification rules and Markdown formatting in STEM manuscripts.

Community
Intermediate
solo-istsolo-ist
2

pipeline-orchestrate

Routes PRs to auto-fix, hold-back or escalate verdicts based on scored risks.

Official
Advanced
gztcodegztcode

requesting-code-review

Automate code review with static security scanning, quality gates, and an independent reviewer subagent.

Community
Advanced
duc01226duc01226

free-tool-strategy

Plan and evaluate free tools for lead generation, SEO, and brand awareness.

Community
Advanced