Store Audit

Store audit → fix → measure loop

Audits the user's whole catalog, compares against saved baselines, and feeds the other claude-fba skills with prioritized findings.

Phase 0 — Preconditions

1. Helium 10 MCP connected (OAuth; mcp__helium10-mcp__* tools available). A paid

Helium 10 plan is required for meaningful data.

1. Check quota first: get_mcp_usage_info (a full audit uses ~20–25 of the 1,000

calls/period).

1. Get the catalog: prefer the user's own cached inventory data or SP-API FBA

inventory (SPAPI.fba_inventory in scripts/sp_api.py). Note: Helium 10's list_my_products requires the Profits module to be ingesting — it can return empty even on healthy accounts; don't treat that as "no products".

Phase 1 — Data pull (~20 MCP calls for a 10–20 ASIN catalog)

• get_listing_score(main_asin=, competitor_asins=[up to 10 own ASINs])

— ONE call returns the 13-point LQS audit for 11 listings.

• get_top_keywords(main_asin=X) per variation-family head — top-10 keyword counts,

captured search volume, per-keyword organic positions. Siblings in a variation family return identical family-level data; don't waste calls on them.

• get_listing_details(main_asin=X) for the top ~6 sellers — sub-category BSR,

reviews, rating, price, estimated sales, 30-day BSR history.

• get_search_query_performance(...) if Brand Registered — the impression→click→

cart→purchase funnel vs. the market per query. First-ever call may return DATAINPROGRESS (retry ~24h).

• list_tracked_keywords / get_keywords_rank_history — only if the user has

populated Helium 10 Keyword Tracker (no API write exists; setup is manual in the UI).

Phase 2 — Diff & report

1. Compare against the newest data/baseline-.json if one exists.
2. Flag: sub-BSR moves >20 positions, keyword rank drops >5, review/rating changes,

LQS regressions, week-over-week conversion shifts >2pts (from track_impact.py snapshots), days-of-cover >180 with Q4 storage-fee season approaching.

1. Write a dated report + save a new baseline JSON with an interventions array —

record every change made and its expected effect so the NEXT audit can attribute.

Phase 3 — Act (use the sibling skills)

• Copy/keyword fixes → listing-guard (draft → user approval → PATCH → backup).
• Review velocity → review-automation (backfill + cron health check).
• Measurement → impact-tracker.
• New/expanded keyword targets → keyword-research (gap analysis, CPR push list).
• Everything not API-executable (bullets on restricted product types, coupons,

removal orders, tracker setup) → a dated manual-actions list for the user.

Phase 4 — Data-driven iteration

Once ≥2 baselines exist, recommend from observed deltas (what actually moved conversion/rank/reviews), not point-in-time heuristics. Scale what worked; kill what didn't.

Hard-won gotchas

• Every ASIN may have 2+ SKUs (FBA + FBM offers) — listing-content PATCHes go to

the BUYABLE one.

• Some product types (e.g. ELECTROSHOCK_WEAPON) have attributes REMOVED from the

schema (bullet_point!) — live content becomes legacy data no API can edit. Check the Product Type Definitions API before promising a fix.

• CPR (units needed in 8 days to reach page-1 top half) under ~15 marks a cheap

ranking push; surface those explicitly.

• Empty generic_keyword (backend search terms) is a free, common win — check it

via the Catalog Items API on every audit.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: EbrahimElbagory
• Source: EbrahimElbagory/claude-fba
• License: MIT

Install and usage instructions live in the source repository linked above.