A forensic validation skill for Claude. Strips optimism bias, corporate fluff, and unverified assumptions from any status report, strategy doc, proposal, or roadmap — and replaces them with what the data actually supports.
Paste any text — a project update, a product strategy, a vendor proposal — and The Skeptic audits it across five dimensions:
- Reality Check Table — compares declared state vs. skeptic-adjusted state with risk ratings
- Fact vs. Assumption Ledger — separates hard facts from leaps of faith
- Domino Effect chains — maps how foundational failures break downstream promises
- Skeptic Verdict — the fatal flaw in two blunt sentences
- Hard Validation Questions — exactly what to ask your team or vendor to demand receipts
Download skeptic-audit.skill and install via Claude Settings → Skills → Install Skill.
- Type
/skepticfollowed by any text - Or paste a status report, proposal, or strategy doc — Claude will recognize when adversarial validation is warranted
Input:
"Migration is on track for Friday. Security review is taking longer than expected, but the lead dev said they will fix any security bugs live in production post-launch so we don't slip."
Skeptic Output (abbreviated):
| Metric | Claimed State | Skeptic-Adjusted State | Risk |
|---|---|---|---|
| Project Health | On track for Friday | "On track" is sentiment, not a milestone report | 🔥 Critical |
| Production Safety | Bugs fixed post-launch | Live production patching = uncontrolled blast radius; no rollback plan cited | 🔥 Critical |
| Security Gate | Delayed but manageable | Security review incomplete at launch; launch is being used to bypass it | 🔥 Critical |
Domino Effect:
Security review incomplete at launch ──► Unpatched vulnerabilities enter production ──► 🔴 Data breach / compliance violation / emergency rollback
Verdict:
This is not a migration on track — it is a migration that has traded a security gate for a calendar date, with live production as the test environment. The phrase "fix any security bugs live in production" is not a mitigation plan; it is a declaration that the launch will be executed in a known-compromised state.
Input:
"Our new AI feature will reduce customer churn by 20% because users love interactive widgets based on our internal beta testing with 5 power users."
Skeptic Output (abbreviated):
| Metric | Claimed State | Skeptic-Adjusted State | Risk |
|---|---|---|---|
| ROI Certainty | 20% churn reduction | Statistically invalid — derived from 5 self-selected power users | 🔥 Critical |
| Sample Representativeness | Internal beta = valid signal | Power users are the least likely to churn; wrong population entirely | 🔥 Critical |
| Causal Attribution | Feature causes reduction | No causal mechanism; correlation from 5 users is not causation | 🔥 Critical |
Verdict:
A churn reduction claim built on 5 power users is not a data point — it is a wish dressed in percentages. Power users are definitionally the cohort least likely to churn, making them the single worst population from which to project churn impact.
# 🕵️♂️ Skeptic Audit: [Subject]
### 📊 Reality Check Table
### 🔍 Fact vs. Assumption Ledger
### 🎲 The Domino Effect
### 🛑 Skeptic Verdict
### ❓ Hard Validation Questions
Part of the claude-skills collection