Skill458 repo starsupdated 2mo ago
purple-cow-audit
The purple-cow-audit skill evaluates whether a product warrants promotion by scoring its remarkability on seven dimensions (uniqueness, quality, story, word-of-mouth potential, design, problem fit, and trust) using web research and a weighted scoring system. Use this before investing time in creating content or marketing materials for a specific affiliate product, especially when deciding between multiple program options.
Install in Claude Code
Copygit clone --depth 1 https://github.com/Affitor/affiliate-skills /tmp/purple-cow-audit && cp -r /tmp/purple-cow-audit/skills/research/purple-cow-audit ~/.claude/skills/purple-cow-auditThen start a new Claude Code session; the skill loads automatically.
Definition
SKILL.md
# Purple Cow Audit
Quality gate for affiliate marketers: score a product's remarkability 1-10 before promoting it. Based on Seth Godin's Purple Cow — if the product isn't remarkable, no amount of marketing skill will make it convert sustainably. The key question: "Would I recommend this to a friend WITHOUT earning a commission?"
## Stage
S1: Research — Evaluating a program's worthiness IS part of research and discovery. This is a quality gate before you invest time creating content, landing pages, and email sequences.
## When to Use
- User is considering promoting a specific product
- User asks "is this product worth promoting?"
- User wants to evaluate product quality before investing time
- User says "purple cow", "remarkable", "audit", "evaluate", "quality check"
- Before investing time in S2-S5 skills for a specific product
- User has a list of programs from `affiliate-program-search` and needs to pick the best
## Input Schema
```yaml
product: # REQUIRED
name: string # Product name
url: string # Product website
description: string # OPTIONAL — what it does
reward_value: string # OPTIONAL — commission rate
tags: string[] # OPTIONAL — categories
comparison_products: string[] # OPTIONAL — competitors to compare against
# Default: auto-discovered
```
**Chaining from S1 affiliate-program-search**: If run, evaluate the `recommended_program`.
## Workflow
### Step 1: Research the Product
1. `web_search`: `"[product] review 2024 2025"` — find recent reviews
2. `web_search`: `"[product] vs" OR "[product] alternative"` — find competitors
3. `web_search`: `"[product] complaints" OR "[product] problems"` — find issues
4. Check product website for: pricing transparency, unique features, social proof
### Step 2: Score Remarkability
Rate each dimension 1-10:
| Dimension | Question | Weight |
|---|---|---|
| **Uniqueness** | Does it do something no competitor does? | 20% |
| **Quality** | Is it genuinely excellent at its core job? | 20% |
| **Story** | Does using it make you feel/look different? | 15% |
| **Word of mouth** | Would users tell friends unprompted? | 15% |
| **Design** | Is the experience delightful, not just functional? | 10% |
| **Problem fit** | Does it solve a real, painful problem? | 10% |
| **Trust** | Transparent pricing, good support, real social proof? | 10% |
**Composite score** = weighted average (1-10)
### Step 3: Make Recommendation
Based on composite score:
- **8-10: PROMOTE** — This is a Purple Cow. Go all in.
- **6-7: PROMOTE WITH ANGLE** — Good product, needs strong positioning. Identify your unique angle.
- **4-5: CAUTION** — Mediocre product. Only promote if commission is exceptional AND you can add significant value through bonuses.
- **1-3: SKIP** — Not remarkable. Promoting this will damage your reputation. Find an alternative.
### Step 4: Identify Remarkable Angles
For products scoring 6+, identify:
1. The 1-2 features that ARE remarkable (Purple Cow elements)
2. The angles that make it share-worthy
3. The audience segment for whom this IS a Purple Cow (even if not for everyone)
### Step 5: Self-Validation
- [ ] Score is evidence-based (cited reviews, features, data)
- [ ] Recommendation is honest (not inflated by high commission)
- [ ] Remarkable angles are specific (not generic praise)
- [ ] Comparison with competitors is fair
- [ ] The "would I recommend without commission" test was honestly applied
## Output Schema
```yaml
output_schema_version: "1.0.0"
purple_cow_audit:
product_name: string
composite_score: number # 1-10 weighted
recommendation: string # "promote" | "promote_with_angle" | "caution" | "skip"
scores:
uniqueness: number
quality: number
story: number
word_of_mouth: number
design: number
problem_fit: number
trust: number
remarkable_angles: string[] # What makes it a Purple Cow (for 6+)
red_flags: string[] # Concerns identified
alternative_products: string[] # Better options if score < 6
remarkability_score: number # Alias for composite_score (for chaining)
chain_metadata:
skill_slug: "purple-cow-audit"
stage: "research"
timestamp: string
suggested_next:
- "affiliate-program-search"
- "grand-slam-offer"
- "viral-post-writer"
- "monopoly-niche-finder"
```
## Output Format
```
## Purple Cow Audit: [Product Name]
### The Question
Would I recommend [product] to a friend WITHOUT earning a commission?
**Answer:** [Yes/No/With caveats]
### Remarkability Scorecard
| Dimension | Score | Evidence |
|---|---|---|
| Uniqueness | X/10 | [specific evidence] |
| Quality | X/10 | [specific evidence] |
| Story | X/10 | [specific evidence] |
| Word of Mouth | X/10 | [specific evidence] |
| Design | X/10 | [specific evidence] |
| Problem Fit | X/10 | [specific evidence] |
| Trust | X/10 | [specific evidence] |
| **Composite** | **X/10** | |
### Recommendation: [PROMOTE / PROMOTE WITH ANGLE / CAUTION / SKIP]
[Reasoning — 2-3 sentences]
### Remarkable Angles (what to emphasize)
1. [Specific remarkable feature/aspect]
2. [Specific remarkable feature/aspect]
### Red Flags (what to be honest about)
1. [Concern]
2. [Concern]
### If Score < 6: Better Alternatives
- [Alternative 1] — [why it's more remarkable]
- [Alternative 2] — [why it's more remarkable]
```
## Error Handling
- **No product provided**: "Tell me the product name and I'll audit its remarkability. Or run `affiliate-program-search` first."
- **Product is too new/no reviews**: Score based on available data, flag low confidence. "Limited data — revisit this audit in 3 months."
- **User disagrees with score**: "The score is a starting framework. If you have personal experience that changes the picture, tell me and I'll adjust."
- **No alternatives found**: Suggest running `affiliate-program-search` in the same category.
## Examples
**Example 1:** "Is HeyGen worth promoting?