Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/zpoint/vibe-sellernpx agentmods add skills/zpoint/vibe-seller/amazon-adsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/zpoint/vibe-seller/amazon-ads)<a href="https://agentmods.dev/skills/zpoint/vibe-seller/amazon-ads"><img src="https://agentmods.dev/badge/skills/zpoint/vibe-seller/amazon-ads/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/zpoint/vibe-seller/amazon-ads"><img src="https://agentmods.dev/badge/skills/zpoint/vibe-seller/amazon-ads.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 70 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00142 | $0.02178 |
| Opus 5 | $0.00071 | $0.01089 |
| Sonnet 5 | $0.00028 | $0.00436 |
| Haiku 4.5 | $0.00014 | $0.00218 |
Grade A, and why
amazon-ads scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Amazon Ads — Catalog
PREREQUISITE: Read
../amazon-shared/SKILL.mdfor marketplace TLD map, hamburger-menu navigation, sign-in / Ziniao / OTP handling, and the ad-console vs seller-central account caveat.
This skill is a catalog. The actual content lives in topical
references in references/. Load whichever ones apply to the task.
What this skill produces
For tuning / audit tasks ("review the ads", "improve ACOS", "audit"):
one Markdown report (AD_AUDIT_<YYYY-MM-DD>.md) + two TSVs per
active campaign (stores/<slug>/ads/<platform>/<country>/<id>.tsv
targets + <id>.searchterms.tsv full customer-query set). Every
campaign is drilled in TWO layers on the same date window — targets AND
search terms — proven by a 搜索词对账 reconciliation line the server
reviewer parses.
START HERE — do NOT pre-read every reference (it buries the model and causes shortcutting). Just two files, then run:
output-spec.md— the report contract (what "done" looks like).audit-quickref.md— the entire procedure on one page. Load a heavy reference only when a step there tells you to.
Then write the report and call
vibe_seller_set_task_result("./AD_AUDIT_<date>.md"). The server's
completeness reviewer replies with a short "what's still missing"
list (under-drilled countries + bid-rule violations) and converges over
rounds — partial is accepted each round, just fix the top gaps and
re-submit until it returns nothing. No separate reviewer subagent or
Stop-hook needed.
Workflow references — the "what to do" thinking
| Reference | Load when |
|---|---|
output-spec.md |
Read first for every audit. The report contract the server completeness reviewer checks against — per-(platform,country) 进度 line, header table, per-campaign drills, the 4 bid rules, TSV-per-campaign. |
audit-quickref.md |
The procedure, one page. Run this top-to-bottom; it points to heavy refs on demand. |
format-anchor.md |
Legacy detail. Per-campaign table column shape; load only if you need the exact table layout. (The mandatory subagent reviewer-loop is superseded by the server completeness reviewer — partial is accepted, it lists gaps each round.) |
reviewer-loop.md |
Phase-4 execution review only (EXEC_REVIEW_*, Stop-hook enforced). For audit reports its Phase-3 format loop is superseded by the server completeness reviewer — don't spawn a review subagent for amazon/noon audits. |
tuning-workflow.md |
User asks to tune ads, improve ACOS, "review last month's ads", harvest search terms, lower bids on losers, weekly ad review, "why is X campaign burning money", or any ongoing-campaign refinement task. |
tuning-campaign-types.md |
A campaign isn't SP-Manual-Keyword. The skill defaults to SP-Manual-Keyword; for SP-Auto / SP-Manual-Product / Sponsored Brands / Sponsored Brands Video / Sponsored Display, this reference has the per-type sidebar tabs, Targeting-tab columns, and lever-applicability matrix observed on a live merchant account. Pair with tuning-workflow.md Phase 3 — that phase branches on type. |
tuning-thresholds.md |
Need to derive per-store thresholds (breakeven ACOS = margin %, target ACOS = 0.7 × breakeven, protect-zone, waste/harvest cutoffs). Always heuristic, never hardcoded. |
tuning-toolbox.md |
Picking the right lever — 8 levers + 2 advanced (dayparting, structural splits) disabled by default. Ordered surgical-first (search-term negate / harvest, per-keyword bid trim) → blanket-last (bidding strategy, pause campaign). For which levers apply per type, see tuning-campaign-types.md. |
tuning-funnel-diagnosis.md |
Distinguishing listing-side problems (low CTR = image / title; low CVR = PDP / price / reviews) from ad-side problems (ACOS) before reaching for a bid lever. Bad CTR is not an ad-tuning problem. |
tuning-recommendation-format.md |
Composing the per-campaign output table at the end of a tuning session — header table → per-campaign data → per-problem subsections with per-entity data tables. Targeting-first, placement-second. Data table shape varies by type — see tuning-campaign-types.md. |
tuning-history.md |
The per-campaign TSV under git that records every observed state (bid, status, suggested range, recent metrics) across audits. Read at Phase 2 to diff scrape against record (catches OBSERVED_DRIFT); read at Phase 3 for recency check before recommending changes (< 7 days since last change → downgrade to Hold); written at Phase 4 after each verified apply (cause-and-effect ledger). One TSV per campaign, written via vibe_seller_write_workspace_file; the workspace auto-commits. |
What ships with it
15 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/audit-quickref.md 9.8 KB
- references/bulk-operations.md 12 KB
- references/format-anchor.md 29 KB
- references/mechanics.md 121 KB
- references/output-spec.md 14 KB
- references/reviewer-loop.md 23 KB
- references/tuning-campaign-types.md 13 KB
- references/tuning-funnel-diagnosis.md 7.5 KB
- references/tuning-history.md 13 KB
- references/tuning-recommendation-format.md 32 KB
- references/tuning-thresholds.md 12 KB
- references/tuning-toolbox.md 16 KB
- references/tuning-workflow.md 42 KB
- requirements.txt 14 B
- scripts/ads_bulk.py 20 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 109 lines · 142 tokens per session scan A 5d2a52cdccd3
amazon-ads is a skill published in the GitHub repository zpoint/vibe-seller (64 stars, last pushed yesterday), licensed Apache-2.0. It adds 142 tokens to every session and 2,178 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
amazon-reviews-api-skill
This skill helps users automatically extract Amazon product reviews via the Amazon Reviews API. Agent should proactively apply this skill when users express needs like getting reviews for Amazon product with ASIN B07TS6R1SF, analyzing customer feedback for a specific Amazon item, getting ratings and comments for a…
amazon-competitor-analyzer
Scrapes Amazon product data from ASINs using browseract.com automation API and performs surgical competitive analysis. Compares specifications, pricing, review quality, and visual strategies to identify competitor moats and vulnerabilities.
asc-subscription-localization
Bulk-localize subscription, subscription-group, and in-app purchase display names across App Store locales using asc, including API 4.4.1 version-scoped v2 resources. Use when filling or updating subscription/IAP names and descriptions without App Store Connect UI work.
food-order
Reorder previous Foodora orders, preview cart contents, and track delivery ETA/status with ordercli. Use when the user wants to reorder food, check delivery status, or browse recent Foodora order history. Never confirm an order without explicit user approval.
product-description-generator
E-commerce product description generator for any platform. Generates optimized titles, bullet points, descriptions, and backend keywords using competitor research + keyword scoring + FABE copywriting. Two modes: (A) Create — generate listing from product specs with optional competitor analysis, (B) Optimize — improve…
amazon-price-tracker
Amazon price monitoring and competitive pricing intelligence. Real-time price tracking, Buy Box analysis, promotion detection, and dynamic pricing strategy optimization. Use when the user asks about price monitoring, competitor pricing, Buy Box tracking, or pricing strategy.