Ab Test Generator

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

Audit 0.4.1 · source sha 2515ce9634da · 15 Oct 2026Report an issue or request a re-scan

• PublisherAffitor GitHub organizationOwner of the repository this skill is read from
• Source folderAffitor/affiliate-skills/skills/analytics/ab-test-generatorThe exact folder we audited
• LicenceMITAs the repository states it
• Commit pinned for this verdict2515ce9634daA new audit runs when the source changes
• Repository statusNot archivedLast push 2 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Generate A/B test variants for affiliate content. Triggers on: "create A/B test", "test my headline", "optimize my CTA", "generate variants", "split test ideas", "improve click-through rate", "test my landing page…

---
name: ab-test-generator
description: >
  Generate A/B test variants for affiliate content. Triggers on:
  "create A/B test", "test my headline", "optimize my CTA", "generate variants",
  "split test ideas", "improve click-through rate", "test my landing page copy",
  "headline alternatives", "CTA variations", "which version is better",
  "optimize conversions", "test my email subject line", "compare approaches".
license: MIT
version: "1.0.0"
tags: ["affiliate-marketing", "analytics", "optimization", "tracking", "ab-testing", "experiments"]
compatibility: "Claude Code, ChatGPT, Gemini CLI, Cursor, Windsurf, OpenClaw, any AI agent"
metadata:
  author: affitor
  version: "1.0"
  stage: S6-Analytics
---

# A/B Test Generator

Generate A/B test variants for affiliate content — headlines, CTAs, landing page sections, email subject lines, and social post hooks. Each variant includes a hypothesis explaining why it might outperform the original. Output is a Markdown document with the original, variants, hypotheses, and a test plan.

## Stage

S6: Analytics — Small changes in headlines and CTAs can swing conversion rates by 20-50%. A/B testing is how professional affiliates systematically find what converts best. This skill removes the guesswork by generating theory-driven variants using proven copywriting frameworks.

## When to Use

- User wants to improve conversion rates on existing content
- User has a headline, CTA, or email subject line and wants alternatives
- User says "test my headline", "optimize my CTA", "A/B test ideas"
- User has a landing page section that isn't converting
- User wants to compare different messaging approaches
- Chaining from S2-S5: take any content output and generate test variants

## Input Schema

```yaml
original: string               # REQUIRED — the content to test (headline, CTA, paragraph,
                               # email subject line, or full social post)

content_type: string           # REQUIRED — "headline" | "cta" | "landing_section"
                               # | "email_subject" | "social_hook"

goal: string                   # OPTIONAL — "clicks" | "signups" | "purchases"
                               # Default: "clicks"

num_variants: number           # OPTIONAL — number of variants to generate (2-5)
                               # Default: 3

audience: string               # OPTIONAL — who sees this content
                               # (e.g., "SaaS founders", "content creators")

product: string                # OPTIONAL — product being promoted
```

**Chaining context**: If S2-S5 content exists in conversation, the user can reference it: "test the headline from my blog post" or "generate CTA variants for my landing page."

## Workflow

### Step 1: Analyze Original Content

Break down the original into components:
- **Emotional angle**: What emotion does it trigger? (curiosity, fear, desire, urgency)
- **Specificity**: How specific vs vague?
- **Structure**: Question, statement, command, statistic?
- **Framework**: Which copywriting framework does it follow? (PAS, AIDA, 4U, BAB)

### Step 2: Identify Testable Elements

Determine what to vary:
- Emotional angle (switch from curiosity to urgency)
- Specificity (add numbers, remove vagueness)
- Structure (question vs statement)
- Length (shorter vs longer)
- Power words (swap key words for stronger alternatives)
- Social proof (add or remove)

### Step 3: Generate Variants

Create `num_variants` alternatives, each using a different approach:
- **Variant A**: Different emotional angle
- **Variant B**: Different structure/format
- **Variant C**: Different specificity level
- Additional variants explore social proof, urgency, or contrarian angles

Each variant must:
- Preserve the core message and product reference
- Preserve any FTC disclosure from the original
- Be a realistic alternative (not just a word swap)

### Step 4: Write Hypotheses

For each variant, explain:
- What was changed and why
- Which copywriting principle supports the change
- What behavior change is expected (e.g., "Higher CTR because questions create open loops")

### Step 5: Suggest Test Plan

Recommend:
- Sample size needed (minimum 100 impressions per variant for social, 500 for landing pages)
- Test duration (7-14 days minimum)
- What metric to track (CTR, conversion rate, revenue per visitor)
- When to declare a winner (95% statistical significance or practical significance threshold)

### Step 6: Self-Validation

Before presenting output, verify:

- [ ] 3-5 distinct variants generated (not just word swaps)
- [ ] Each hypothesis grounded in a copywriting principle or framework
- [ ] Sample size calculation is present and realistic
- [ ] Test duration is ≥7 days minimum
- [ ] Winner criteria defined with statistical significance threshold

If any check fails, fix the output before delivering. Do not flag the checklist to the user — just ensure the output passes.

## Output Schema

```yaml
output_schema_version: "1.0.0"  # Semver — bump major on breaking changes
test:
  original: string
  content_type: string
  goal: string

variants:
  - label: string              # "Variant A", "Variant B", etc.
    content: string            # the variant text
    change: string             # what was changed
    framework: string          # copywriting principle used
    hypothesis: string         # why this might win

test_plan:
  sample_size: number          # per variant
  duration: string             # recommended test period
  metric: string               # what to measure
  winner_criteria: string      # when to pick a winner
```

## Output Format

1. **Original** — the current content being tested
2. **Variants** — each variant with its content, change description, and hypothesis
3. **Test Plan** — sample size, duration, metric, winner criteria
4. **Quick Win** — if one variant is clearly stronger based on copywriting principles, call it out

## Error Handling

- **Original too short (1-2 words)**: "I need more cont

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

1. 2026-10-15
   ✓SAFE B 89 / 100

   first audit · source 2515ce9634da

Questions about Ab Test Generator

• github.com/Affitor/affiliate-skills/tree/main/skills/analytics/a — the audited source at commit 2515ce9634da and every finding aboveObserved · 15 Oct 2026

Spotted something wrong? Report a correction · Methodology