Amazon PPC Advanced Skills

AI assistant workflows for Amazon PPC diagnosis, growth planning, search-term harvesting, product-aware Rocketcart MCP reviews, masked optimization output, and approval-ready action queues.

Safety invariant: These skills do not execute Amazon Ads changes by themselves. Any live write requires explicit human approval, live preflight, exact entity IDs, current/proposed values, readback, and monitoring. If any requirement is missing, the action is not executable.

Start Here

What This Repo Is

• A library of standalone Codex or Claude skills for Amazon PPC work.
• A set of examples, eval prompts, and stress tests for safe agent behavior.
• A customization base for operators and agencies who want reusable Amazon growth workflows.
• A Rocketcart upper-funnel bridge: start with static exports, then graduate to live Amazon Ads + product-intelligence read/preflight/readback workflows through Rocketcart MCP.

What This Repo Is Not

• Not an Amazon Ads API connector by itself.
• Not an automatic bid, budget, negative, placement, or campaign changer.
• Not a replacement for human approval on live account mutations.
• Not a place to commit real Amazon, Rocketcart, customer, or proprietary account data.
• Not a public home for masking registries, source-ID mappings, raw reports, credentials, private execution manifests, or customer-specific codenames.

Standalone Skills Vs Rocketcart MCP

Rocketcart MCP is not just an Amazon Ads connector. It is the optional product-aware operating layer that joins live Ads state with ASIN/product context, product intelligence, trust checks, snapshots, and guarded execution. See Rocketcart MCP guide.

Rocketcart-Aware Review Modes

Use rocketcart-amazon-ads-live-review for broad live account, product-aware growth, preflight, and monitoring reviews. Use amazon-search-term-harvest-planner directly when the Rocketcart job is specifically search-term harvesting, source-negative routing, product-target expansion, or harvest-row execution.

Try The Rocketcart Bridge In 60 Seconds

Codex:

git clone https://github.com/nospicyplease/amazon-ppc-advanced-skills.git
cd amazon-ppc-advanced-skills
mkdir -p ~/.codex/skills
cp -R rocketcart-amazon-ads-live-review ~/.codex/skills/

Claude:

zip -r rocketcart-amazon-ads-live-review.zip rocketcart-amazon-ads-live-review

Upload that one ZIP as a Claude skill. Do not upload the whole repo as one skill.

First smoke test, even without Rocketcart connected:

Use $rocketcart-amazon-ads-live-review in Live Optimization Review mode. I do not have Rocketcart MCP connected yet. Run in standalone mode and tell me the exact Amazon Ads and product data you need for a safe first review. Do not execute anything.

Expected first output shape:

Mode: Standalone
Review mode: Live Optimization Review
Execution: no writes; approval required for any future live action
Missing data: campaigns, search terms, product/ASIN context, inventory, offer status, margin, BSR/category movement, competitor signals, recent changes
Next step: provide exports or connect Rocketcart MCP for live reads

With Rocketcart MCP connected, paste one of these:

Use $rocketcart-amazon-ads-live-review in Live Optimization Review mode for profile example_de. Confirm the profile, inspect live Sponsored Products campaigns, product ads/ASIN mapping, budget and targeting drift, snapshots/changelogs, product readiness, category/BSR movement, and competitor signals. Produce proposed action rows only. Do not execute anything.

Use $rocketcart-amazon-ads-live-review in Product-Aware Growth Review mode for profile example_de. Classify each ASIN/campaign as Grow, Fix Before Scaling, Protect, Monitor, or Blocked using Ads performance plus inventory, Featured Offer / Buy Box, price, reviews/rating, estimated demand, category/BSR movement, competitor signals, margin, and recent changes. Do not execute anything.

Use $rocketcart-amazon-ads-live-review in Preflight / Approval Readiness Review mode for profile example_de. Review these candidate action rows for exact entity IDs, current values, proposed values, product-readiness gates, expected impact, risk, approval text, readback, and monitoring. Mark each row Approval Ready, Needs IDs, Needs Current Value, Needs Product Context, Stale Approval, Blocked, or Monitor Only. Do not execute anything.

Use $rocketcart-amazon-ads-live-review in Post-Change Readback / Monitoring Review mode for profile example_de. Review the approved changes from the last execution window, read back affected entities, compare expected versus current state, and classify each action as Readback Confirmed, Partially Applied, Not Applied, Monitoring, Worked, Failed, or Needs More Data. Do not execute new writes.

Without Rocketcart, provide exports or pasted summaries. Rocketcart adds live campaign state, product-ad/ASIN mapping, product intelligence, recent-change context, preflight, readback, and guarded execution. In both modes, any live write still needs explicit human approval, live preflight, exact IDs, current/proposed values, readback, and monitoring.

First Run Prompt

Use $amazon-account-growth-operating-system to build a weekly Amazon PPC action plan from the attached campaign, search term, Business Report, BSR, inventory, and retail-readiness data. Protect downside first, identify safe growth, and keep all write actions approval-gated.

For a first test with one search term CSV:

Use $amazon-search-term-harvest-planner with this search term report. Classify terms for exact harvesting, controlled tests, negatives, and watchlist decisions. Do not execute anything.

For a Rocketcart search-term harvest review:

Use $amazon-search-term-harvest-planner in Live Harvest Review mode for profile example_de. Resolve live campaign/ad group/keyword/negative IDs, check duplicate exacts, current negatives, destination feasibility, product-ad ASIN/SKU context, recent drift, and product readiness. Produce approval-gated harvest rows only. Do not execute anything.

For a first Rocketcart MCP review:

Use $rocketcart-amazon-ads-live-review for profile example_de. Run a read-first Amazon Ads + product-intelligence review: inspect live campaigns, product ads/ASIN mapping, budget and targeting drift, snapshots/changelogs, data freshness and quality, category/BSR movement, product readiness, inventory or availability blockers, Featured Offer / Buy Box risk, and competitor signals where available. Produce proposed action rows only. Do not execute anything.

Minimum Data Checklist

You can start with partial data, but missing data must lower confidence. Useful inputs include:

• Marketplace, profile/account, currency, timezone, ASIN/campaign scope.
• Exact current and comparison windows, preferably T-1 complete.
• Campaign, targeting, search term, placement, advertised-product, and purchased-product reports.
• Business Reports or total sales for TACoS and incrementality.
• BSR history and organic rank when rank or velocity matters.
• Inventory, Featured Offer / Buy Box, price, reviews, rating, listing status, and delivery promise.
• Margin, target ACoS, target CPA, or product economics.
• Recent changes: bids, budgets, placements, negatives, launches, pauses, listings, price, coupons, deals, and inventory events.

Install Quickly

git clone https://github.com/nospicyplease/amazon-ppc-advanced-skills.git
cd amazon-ppc-advanced-skills
mkdir -p ~/.codex/skills
cp -R amazon-account-growth-operating-system ~/.codex/skills/

For Claude, zip and upload one skill folder at a time:

zip -r amazon-account-growth-operating-system.zip amazon-account-growth-operating-system

Do not upload the whole repository to Claude as one skill. See Installation for full instructions.

Skills

1. amazon-ads-performance-drop-diagnosis

   • Diagnoses why Amazon Ads, retail sales, BSR, conversion, TACoS, ROAS, or rank performance declined.
   • Protects sales velocity and organic momentum before recommending bid, budget, negative, pause, or relaunch actions.

2. amazon-growth-opportunity-finder

   • Finds ASIN, campaign, keyword, search term, product-target, placement, BSR, and organic-rank opportunities that can grow profitably.
   • Checks margin, incrementality, retail readiness, Featured Offer / Buy Box, inventory, reviews, conversion, and BSR context before scale.

3. amazon-account-growth-operating-system

   • Orchestrates downside and upside evidence into one prioritized account action queue.
   • Decides what to protect, fix, scale, reduce, harvest, launch, monitor, and approval-gate.

4. amazon-search-term-harvest-planner

   • Finds search terms ready for exact-match harvesting from auto, broad, phrase, or discovery campaigns.
   • Works standalone from exports or in Rocketcart MCP mode for live resolution, preflight, exact row approval, guarded execution, readback, and monitoring.
   • Produces PLANNING_ONLY, NEEDS_DATA, BLOCKED, APPROVAL_REQUIRED, or APPROVAL_READY rows and blocks unsafe source negatives when traffic may be brand defense, own-ASIN defense, launch/rank support, or low-sample discovery.

5. rocketcart-amazon-ads-live-review

   • Runs read-first live optimization, product-aware growth, preflight readiness, and post-change monitoring reviews in standalone or Rocketcart MCP mode.
   • Uses Rocketcart MCP, when available, to inspect profiles, live campaigns, product ads/ASIN mapping, budget and targeting drift, snapshots, changelogs, category/BSR movement, product context, and readiness blockers before classifying ASINs/campaigns as Grow, Fix Before Scaling, Protect, Monitor, or Blocked.

6. case-camouflage-skill

   • Lives under skills/case-camouflage-skill.
   • Keeps analytical-plane optimization logic on raw source IDs and exact KPIs, then masks only user-facing display labels and identifiers.
   • Provides a Python reference package under src/case_camouflage_skill/ for registry providers, masking, metric validation, approval packets, private manifests, scanners, synthetic E2E tests, real-profile dry-run gating, and readiness reporting.
   • Does not mutate Amazon Ads. It may produce masked approval packets and private execution manifests for a separate approved execution tool.

Repository Layout

amazon-*/                         Production skill folders
rocketcart-*/                      Rocketcart-aware skill folders
skills/                            New-style skill folders, including case-camouflage-skill
src/                               Testable helper modules for case-camouflage-skill
schemas/                           JSON schemas for masked packets, registries, manifests, and inputs
tests/                             Unit, leakage, E2E, and gated real-profile dry-run tests
docs/                              Install, FAQ, glossary, workflow, maintenance, data privacy
examples/                          Reproducible fixture packs for every production skill
evals/                             Manual review prompts and concrete eval cases
stress-tests/                      Adversarial prompts and expected resistance behavior
templates/                         New skill scaffold
.github/                           Issue templates, PR template, validation workflow

Examples, Evals, And Stress Tests

Each production skill has an example under examples/:

• prompt.md: realistic anonymized prompt.
• input-summary.md: available fields, missing fields, assumptions, and scope.
• expected-output-outline.md: sections and safety behavior a good answer should include.
• sample-data/: synthetic CSV or JSON fixtures.
• expected-output.md: concrete good-answer target.
• known-bad-output.md: unsafe or low-quality answer that should fail review.
• eval-result.md: which eval prompts should pass or fail.

Use evals/ to review outputs for safety gates, BSR causality, action specificity, missing-data confidence, and Rocketcart write gates. Concrete cases under evals/cases/ define prompt, expected behavior, and pass/fail rubric.

Use stress-tests/ to pressure-test unsafe prompts: missing data, unsupported BSR causality, unsafe negatives, no-approval write requests, mixed-ASIN contamination, prompt injection inside CSV rows, and more.

Masked Optimization Output Commands

The masked-output package uses synthetic fixtures by default and does not require real credentials:

make test
make test-unit
make test-e2e
make test-leakage
make test-real-dry-run
make production-readiness

make production-readiness reports NOT_PRODUCTION_READY unless synthetic tests, leakage tests, and gated read-only real-profile dry-runs pass. Live execution tests remain disabled unless ALLOW_LIVE_EXECUTION_TESTS=true, an allowlist, and a separate execution adapter are configured.

For real-profile dry-runs, use ignored private paths only:

private-test-data/registry.json
private-test-data/profile-<profile_id>.json

Create the ignored local scaffold with:

make prepare-private-dry-run
make check-private-dry-run-config

Set ALLOW_REAL_PROFILE_TESTS=true, PRIVATE_TEST_DATA_DIR, MASKING_REGISTRY_URI, MASKING_HMAC_SECRET, and Amazon Ads credential/profile env vars in your shell or ignored .env. Profile fixtures must use fixture_kind: real_profile_read_only; the harness will block placeholder-looking sensitive env values and non-private fixture paths.

Run the structural checks before opening a PR:

make check-docs
make check-examples
make review-fixtures
make validate
make eval

make review-fixtures is the contributor quality gate for examples, eval cases, stress tests, and unsafe write-language checks. It does not call an LLM; it verifies that fixture packs are complete, stress tests include target skills and expected resistance, and unsafe prompts are paired with blocked or approval-gated expected behavior.

Documentation

• Installation
• FAQ
• Glossary
• Skill catalog
• Operating workflow
• Rocketcart MCP guide
• Data privacy
• Maintenance and update guide
• Contributing
• Roadmap
• Examples
• Evaluation prompts
• Stress tests

Contributing

Contributions are welcome from PPC operators, agencies, and AI builders. Start with CONTRIBUTING.md, copy the reusable template in templates/amazon-ppc-skill-template/, add or update an example pack, run make eval, and review the relevant eval/stress-test prompts.

License

MIT. See LICENSE.