| name | cvr-leak-fix |
| description | End-to-end CVR-leak fix pipeline. Detects keyword × visual coverage gaps via SellerApp, generates the missing visual via Higgsfield, validates with ProductPinion poll, outputs ship-ready slot. Use when running `cvr-leak-fix {ASIN}` or when CVR is below category average and you need a single-slot intervention. Closes the loop from diagnosis to ship-ready in one pipeline. |
cvr-leak-fix — Full CVR-Leak Diagnostic-to-Ship Pipeline
The high-leverage CVR play: find the biggest keyword × visual gap, generate the visual that fills it, validate with shoppers. One slot, three tools, ship-ready output.
Prerequisites
- SellerApp via n8n MCP
- Higgsfield MCP
- ProductPinion MCP
Invocation
cvr-leak-fix {ASIN}
cvr-leak-fix {ASIN} --leak-rank=1 # default: fix top leak; use 2/3 for next-priority
Output
/tmp/cro-content/{ASIN}-cvr-leak-fix-{date}.md + ship-ready image at /tmp/cro-content/{ASIN}-cvr-fix-WINNER.png.
State File
/tmp/cro-content/{ASIN}-cvr-leak-fix.state.json for resumable runs.
Phase 1 — Detect
Calls keyword-cvr-leak {ASIN} → top-10 leaks with recommended fixes.
Phase 2 — Pick the Target
Default: top-ranked leak. User can override with --leak-rank=N.
The leak record contains:
- Keyword + search volume
- Implied benefit
- Coverage gap (image / bullet / title)
- Recommended fix type
Phase 3 — Generate the Fix
Branch by fix type:
- Image gap →
infographic-builder for benefit overlays OR lifestyle-stack-generator for use-context
- Bullet gap →
copy-split-test with 3 bullet variants generated
- Title gap →
copy-split-test with 3 title variants
Skill picks the right downstream skill, passes the leak's benefit constraint, gets the candidate output.
Phase 4 — Validate
Branch by fix type:
- Image fix →
main-image-poll if main, three-second-test if hero, or stacked-gallery-test if it changes the whole stack
- Copy fix →
copy-split-test (ProductPinion text split)
Use existing brand audience via audience-builder lookup.
Phase 5 — Output (Ship-Ready)
# CVR-Leak Fix — {ASIN}
**Leak fixed:** {keyword} ({vol}/mo) — benefit "{benefit}" was searched but not visually shown
**Date:** {date}
## Pipeline Trace
| Phase | Output | Link |
|-------|--------|------|
| 1. Detect | Leak rank #{N} | [keyword-cvr-leak] |
| 2. Generate | {N candidates} | [generation output] |
| 3. Validate | Winner: {variant} at {N}% preference | [poll output] |
## Winner

**Why this fixes the leak:**
{trace from leak to fix to validation}
## MYE Launch Plan
- Test: New {slot} vs current
- Duration: 14 days
- Success: CVR lift ≥10% for visitors searching the leak keyword
- Source-attribution: SQP keyword-level CVR (track {keyword} specifically)
## Recommended Next Skill
- After 14 days: pull SQP, validate the leak closed → run `cvr-leak-fix {ASIN} --leak-rank=2` for next-priority leak
Reference Files
~/.claude/knowledge/cro-methodology/decision-framework.md (CVR > CTR priority)
- Vault:
CRO-Knowledge-Base/MASTER-CRO-REFERENCE.md §4 (CVR-leak rules)
Quality Bar
Auto-Triggers
- User says "fix CVR for {ASIN}" / "what's leaking on {ASIN}"
- Diagnostic shows healthy CTR but low CVR
- Following
/cro-priority-tracker or sqp-priority-rank