Skip to main content
skill CX Codex

dark-pattern-audit — Claude Code skill: dark pattern audit for code

dark-pattern-audit reads source files and UI copy for manipulative design: preselected boxes, hidden fees, fake countdowns. Nothing needs to run.

MIT 219 installs ★ 35 updated 2026-10-21

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add podo/design-agent-skills --skill dark-pattern-audit -a claude-code
Open on GitHub

What dark-pattern-audit does

The skill works from local files only: it reads source code and interface copy and looks for manipulative patterns such as preselected boxes, fees added late in checkout, countdown timers, and CTA labels that hide the real action. No screenshots, no running app. Coverage runs across pricing and cart, subscription and cancellation screens, consent prompts, modals, and steps a user cannot skip.

Each finding comes back with the pattern name and a link to its reference description, the file, line and component where it sits, plus a short reason drawn from what is in the code. Grading is conservative, so only high-confidence hits make the report; a borderline case stays out. The run goes in four moves: agree on scope, load the pattern catalog, collect evidence, write the report.

Auto-start is off. The agent will not begin an audit on its own, so you call /dark-pattern-audit in Claude or name the skill in your message. Reach for wcag-ai-skill when the job is WCAG conformance, and design-auditor for general design quality.

What it helps with

  • find preselected consent boxes in form markup
  • trace where extra fees get added to the price
  • map the forced steps in a cancellation flow
  • list modals with fake countdowns or stock counters
  • catch CTA copy that hides what the button really does

Example prompts for the agent

  • Run /dark-pattern-audit over src/checkout and tell me where the extra fees come from.
  • Audit the subscription and cancellation flows; I want file and line for every finding.
  • Check the button labels and modal copy for manipulative phrasing, high-confidence findings only.

How to install dark-pattern-audit

1Skills CLI (Vercel)

npx skills add podo/design-agent-skills --skill dark-pattern-audit -a claude-code

2Manually: put the files in place

~/.claude/skills/dark-pattern-audit/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/dark-pattern-audit/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about dark-pattern-audit

How do I install dark-pattern-audit in Claude Code?

Run npx skills add podo/design-agent-skills --skill dark-pattern-audit -a claude-code. The skill file lands at ~/.claude/skills/dark-pattern-audit/SKILL.md.

What does dark-pattern-audit do?

It reads code and interface copy, flags manipulative patterns, and writes a report with the pattern name, a reference link, the exact location, and a short reason. It stays on static evidence, so there is no page to open. Auto-invocation is disabled: call /dark-pattern-audit yourself.

Is this the same as an accessibility audit?

No. WCAG conformance belongs to wcag-ai-skill, and plain design quality to design-auditor; this skill targets deception in the interface.

Source text

SKILL.md · MIT
---
name: dark-pattern-audit
description: |
  Audits codebases and product copy for deceptive dark patterns inferable from
  local source code and text — no runtime inspection or screenshots needed.
  Produces a structured narrative report with confidence grading, source
  locations, concise reasoning, and official pattern reference links for each
  detected pattern. Covers pricing, subscriptions, consent, modals, preselected
  controls, misleading CTAs, and more.
triggers:
  - "dark pattern audit"
  - "dark pattern detection"
  - "deceptive design audit"
  - "dark-pattern-audit"
  - "deceptive patterns"
  - "manipulative UI audit"
das:
  type: skill
  category: accessibility-quality
  upstream: "https://github.com/SidKH/skills"
  upstream_path: "dark-pattern-audit/SKILL.md"
  version: latest
---

# dark-pattern-audit

> Catalogue stub — full skill: [SidKH/skills](https://github.com/SidKH/skills/tree/main/dark-pattern-audit)

<!-- das:install:start — generated from stub.yaml by `npm run generate`; edit stub.yaml, not this block -->
## Decision tree

1. **Is the full skill already installed?**
   - Global: `f=~/.agents/skills/dark-pattern-audit/SKILL.md; [ -f "$f" ] && ! grep -q "^das:" "$f" && echo "installed" || echo "pointer"`
   - Project: `f=.agents/skills/dark-pattern-audit/SKILL.md; [ -f "$f" ] && ! grep -q "^das:" "$f" && echo "installed" || echo "pointer"`
   - `installed` → the skill is installed as `dark-pattern-audit`, invoke it and proceed
   - `pointer` → go to step 2

2. **Detect scope, then install:**

   ```bash
   [ -e ~/.agents/skills/dark-pattern-audit ] && echo "global" || echo "project"
   ```

   **Global:**
   ```bash
   npx skills add SidKH/skills --skill dark-pattern-audit -g -y
   ```

   **Project:**
   ```bash
   npx skills add SidKH/skills --skill dark-pattern-audit -y
   ```
   > **Claude Code:** send a command as a chat message starting with `!` to run it without leaving the conversation.

**Heads-up**

- **License:** no LICENSE file in the repo
<!-- das:install:end -->

## Invoke after install

- Skill name: `dark-pattern-audit`
- The upstream skill sets `disable-model-invocation: true`: an agent will not start it on its own. Ask the user to run `/dark-pattern-audit` (in Claude Code), or to name the skill explicitly in the prompt.
- Trigger phrases: "dark pattern audit", "dark pattern detection", "deceptive design audit"

## What it does

Static analysis of a codebase for dark patterns — purely from source code and copy, without runtime access:

**Surfaces covered:**
- Pricing, checkout, billing, fees, cart, upsells
- Subscriptions, trials, renewals, cancellation flows
- Consent dialogs, privacy prompts, notification permissions
- Modals, interstitials, countdown timers, fake stock counters
- Preselected controls, bundled choices, forced steps
- Misleading CTA text, low-contrast disclosures, hidden disclaimers

**Output per finding:**
- Pattern name + official reference link
- Confidence grade (conservative — only high-confidence findings reported)
- Exact source location (file, line, component)
- Concise reasoning grounded in local evidence

**Workflow:** Confirm scope → load pattern catalog → search evidence in code/copy → report

## When NOT to use

- General design quality audit (typography, dark mode, RTL) → use `design-auditor`
- WCAG accessibility compliance → use `wcag-ai-skill`
- AI writing pattern detection → use `humanize-text`
- Ethical design strategy → use `plan-design-review`

© podo · MIT · podo/design-agent-skills

Similar skills

All skills for Design →

From the same set108

The whole set →
cloudflare-web-perf cloudflare-web-perf audits Core Web Vitals, render-blocking assets, and Lighthouse performance scores. It diagnoses and fixes LCP, INP, and CLS bottlenecks using Chrome DevTools traces. CX Codex +1 221installs design-catalogue design-catalogue finds the right design skill from a catalogue, maps it to the user’s goal, and points to installation and invocation steps. CX Codex +1 221installs design-review-garrytan design-review-garrytan runs a visual audit of your UI, then lands each fix as its own atomic commit. Useful when design debt has piled up and you want it cleared in reviewable pieces. CX Codex +1 221installs bencium-ux-designer bencium-ux-designer — a design-direction skill: it drafts ten typography-only HTML directions, waits for your pick, then builds the brand or product system from it. CX Codex +1 221installs make-interfaces-better make-interfaces-better — a skill for Claude Code, Cursor, Codex and other platforms that teaches the agent to make interfaces feel genuinely alive. Applies Jakub Krehel's design engineering principles — creator of oklch.fyi and Interfaces. CX Codex +1 219installs designer-skills designer-skills is a Claude Code skill with 111 skills across nine plugins, covering design work from user research to visual critique. CX Codex +1 218installs neo-user-journey neo-user-journey maps end-to-end user journeys with emotional tracking and tests live flows using Playwright. Useful for auditing interactive web apps against UX heuristics. CX Codex +1 217installs guizang-ppt guizang-ppt turns an article, outline, or Markdown file into a single HTML web deck. Use it for swipe navigation, presenter notes, and rehearsal screens. CX Codex +1 217installs

Project sponsors

Companies that keep this analytics open