Communitygithub.com

peterfei/landing-page-skill

Full-stack landing page generation Agent Skill. Trigger words: landing page, product page, marketing page, homepage design. 5-phase protocol: Requirements Gathering → Strategy Planning → Design System → Code Generation → Quality Validation. Suitable for single-page conversion landing pages (SaaS/e-commerce/services). Not suitable for multi-page sites, backend integration, or complex interactive applications.

What is landing-page-skill?

This agent skill, `peterfei/landing-page-skill`, provides a complete, structured workflow for generating high-converting landing pages. It follows a rigorous 5-phase protocol covering requirements gathering, strategy planning, design system definition, code generation (Next.js 14/15 + Tailwind CSS), and automated quality validation. The skill is agent-agnostic, compatible with Claude Code, Codex CLI, Cursor, Trae, OpenClaw, Hermes Agent, and other markdown-based skill runtimes. It includes industry templates, copywriting formulas, aesthetic styles, anti-pattern detection, and a self-healing loop for blocking issues, ensuring both conversion optimization and code quality from the start. The skill is ideal for developers and AI agent users who need to rapidly prototype high-quality, single-page landing pages without writing boilerplate from scratch.

Works with✓Claude Code✓Codex CLI✓Cursor
npx skills add peterfei/landing-page-skill

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Landing Page Skill

This Skill is designed to be agent-agnostic — usable with Claude Code, Codex CLI, Cursor, Trae, OpenClaw, Hermes Agent, or any agent supporting markdown-based skills. All file references use relative paths from this Skill's root directory.

Activation Conditions

Activate when user input contains any of the following trigger words: landing page, product page, marketing page, homepage design, single page website.

Honesty Boundaries (Applicability)

Applicable:

  1. Single-page conversion landing pages (collecting leads, guiding sign-ups, driving purchases)
  2. Product/service showcase pages (SaaS, single e-commerce items, consulting services, courses)
  3. Tech stack: Next.js 14/15 + Tailwind CSS (MVP stage)
  4. Pages with a clear conversion goal (CTA)

Not Applicable:

  1. Multi-page websites or blog systems (beyond single-page scope)
  2. Full applications requiring backend API, database, or user authentication
  3. Complex interactions (real-time collaboration, data visualization dashboards, games)
  4. Non-web platforms (mini-programs, native apps, desktop applications)

Boundary Statement: This Skill generates a high-quality starting codebase; manual polishing of copy details and brand assets is required. It does not promise "one-click perfection."

5-Phase Operational Protocol

Phase 1: Requirements Gathering

Confirm the following information with the user (skip if already provided):

  1. Product/Service: What is it? What is the core value proposition?
  2. Target Audience: Who will use it? What pain points do they have?
  3. Conversion Goal: What do you want the user to do? (Sign up / Purchase / Download / Book)
  4. Tech Stack: Confirm use of Next.js 14/15 + Tailwind CSS
  5. Brand Style Preference: Choose from 6 styles in references/aesthetic-styles.md
  6. Competitor References: Any preferred landing page references? (Optional)

Output: Requirements summary document (Markdown List), used in Phase 2.

Phase 2: Strategy Planning

  1. Load Industry Template: Select template from references/page-templates/ based on product type:
    • SaaS Product → references/page-templates/saas-product.md
    • E-commerce Single Item → references/page-templates/ecommerce.md
    • Other → General structure (Hero → Features → Benefits → Social Proof → CTA → Footer)
  2. Select Aesthetic Direction: Choose a style from references/aesthetic-styles.md, confirm color tendencies
  3. Copy Structure: Plan using references/copy-formulas/:
    • Hero Headline: "Result + Audience + Mechanism" formula from references/copy-formulas/hero-headlines.md
    • CTA Copy: "Verb + Benefit" format from references/copy-formulas/cta-patterns.md
    • Objection Handling: 5 objection templates from references/copy-formulas/objection-handling.md
  4. Conversion Framework Mapping: Map 7-part copy to DESIGNNAS 11 Elements (see references/conversion-framework.md)

Output: Strategy document (including style selection, color direction, copy outline, element mapping table).

Phase 3: Design System

Generate design tokens based on references/design-system.md:

  1. Typography Tokens: Display font, Body font, heading size scale
  2. Color Tokens: Primary / Secondary / Accent / Neutral / Background / Text (hex + CSS variables)
  3. Motion Tokens: PageLoad animation, Scroll-triggered animation, Hover interaction (type + duration)
  4. Layout Tokens: Grid type, Spacing scale, Section width, asymmetrical layout rules

Output: Design token document (directly copyable into tailwind.config.ts).

Phase 4: Code Generation

Generate code following references/tech-adapters/nextjs-tailwind.md:

  1. Project Structure: app/page.tsx, app/layout.tsx, components/, lib/utils.ts, tailwind.config.ts
  2. Componentization: Each DESIGNNAS element as an independent component (HeroSection, FeaturesSection, etc.)
  3. Tailwind Configuration: Extend design system tokens (colors, fonts, animations)
  4. Responsive Design: Use sm:/md:/lg:/xl: breakpoints
  5. Accessibility: Semantic HTML, alt text, contrast, focus-visible

Output: Complete runnable Next.js project code.

Phase 5: Quality Validation (Closed-Loop Assurance)

5.1 Conversion Check

  • 11-element coverage: Check for Hero, Social Proof, Features, Benefits, How It Works, Testimonials, Pricing/CTA, FAQ, Scarcity, Footer (at least 7 covered)
  • CTA Clarity: Is CTA text using "Verb + Benefit" format?
  • Hero Headline: Does it follow the "Result + Audience + Mechanism" formula?

5.2 Anti-pattern Detection (Check each item, see references/anti-patterns.md)

#RuleDetection LogicSeverity
1Overused FontInter/Roboto/Arial as Display fontBlocking
2Template Color SchemePurple gradient + pure white background template comboBlocking
3Vague CTACTA is Submit/Learn More/Click HereBlocking
4Missing Social ProofNo Social Proof element below HeroBlocking
5Weak Hero HeadlineStarts with Welcome/Brand name, purely descriptiveBlocking
6Missing Mobile ResponsivenessNo sm:/md:/lg: responsive breakpointsWarning
7Accessibility BaselineMissing alt/aria/semantic HTMLWarning
8Missing SEO BasicsNo meta/title/OG tagsWarning
9Performance RiskLarge images unoptimized, missing lazy loadWarning
10Token ConsistencyHardcoded colors/non-standard spacingWarning

5.3 Auto-fix Loop (Blocking Rules)

When blocking rules (1-5) fail, automatically execute the fix loop:

  1. For each failing blocking rule, generate specific fix code
  2. Replace only the Section code triggering the rule (do not modify unrelated Sections)
  3. Re-run detection for that rule
  4. Repeat the above steps, up to 3 rounds
  5. If still failing after 3 rounds → Report the failure reason, leave decision to user

5.4 Positive Pattern Validation (Advisory, see references/positive-patterns.md)

#PatternDetection Logic
P1CTA Verb+BenefitCTA text matches "Verb + Specific Benefit" format
P2Hero Result+AudienceHeadline contains quantifiable result + target audience
P3Social Proof ElementsPage contains ≥1 social proof type
P4Urgency/ScarcityPresence of spots/time/price urgency elements

Positive patterns do not trigger auto-fix; they only provide optimization suggestions in the report.

Output: Detection report (Anti-pattern + Positive pattern Markdown tables) + Auto-fix summary (if any fixes applied).

Phase 5.5: Iterative Optimization (Optional)

After Phase 5, offer the user 3 options:

A) Accept — Output final complete project code B) Modify specified Section — User describes the Section to modify and the modification request C) Regenerate specified Section — User specifies a Section to regenerate from scratch

Iteration Rules:

  • Modifications/regenerations only affect the user-specified Section (section granularity)
  • After each modification, re-detect anti-pattern rules related to that Section
  • Up to 3 rounds of iteration, output current version after limit exceeded
  • Each round outputs a change log: "Round N: Modified [Sections], Re-checked [Rules]"

Index

Tech Stack Adapters

  • Next.js 14/15 + Tailwind CSS → references/tech-adapters/nextjs-tailwind.md

Industry Templates

  • SaaS Product Page → references/page-templates/saas-product.md
  • E-commerce Product Page → references/page-templates/ecommerce.md

Copy Formulas Library

  • Hero Headlines → references/copy-formulas/hero-headlines.md
  • CTA Patterns → references/copy-formulas/cta-patterns.md
  • Objection Handling → references/copy-formulas/objection-handling.md

Design References

  • Conversion Framework → references/conversion-framework.md
  • Design System → references/design-system.md
  • Aesthetic Styles → references/aesthetic-styles.md
  • Anti-pattern Detection → references/anti-patterns.md
  • Positive Pattern Detection → references/positive-patterns.md

Cross-Agent Environment Adaptation Notes

This Skill is designed to be agent-agnostic, compatible with the following runtimes:

RuntimeSupport MethodNotes
Claude CodeNative supportRecommended, activate via /agent or trigger words
Codex CLICompatibleDirectly load SKILL.md
CursorCompatibleCopy to .cursorrules or use Composer
TraeCompatibleDirectly load SKILL.md
OpenClawCompatibleFollows skills.sh protocol
Hermes AgentCompatibleDirectly load SKILL.md
Other markdown-based skill runtimesCompatibleAny runtime supporting YAML frontmatter + Markdown

Path Convention

All file references use relative paths from this Skill's root directory:

  • references/xxx.md
  • assets/xxx.svg
  • scripts/xxx.py

The agent or user resolves based on their own installation location, without relying on any absolute paths.

Self-Containment Principle

SKILL.md contains all necessary information and can be copied and used directly. Files in references/ provide deep knowledge supplements and can be loaded as needed.

This Skill follows the Agent Skills open protocol.

Related Skills