Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Read and mutate SP, SB, and SD campaigns, ad groups, ads, targeting, and budget rules.
Amazon Ads Manager
Read and mutate SP, SB, and SD campaigns, ad groups, ads, targeting, and budget rules.
The package preserves 61 Amazon-platform operations (18 read or connection operations and 43 mutations). It uses only NEXSCOPE_PROXY_BASE and NEXSCOPE_API_KEY. Provider credentials, refresh, request signing, authorization state, upload locations, and document locations remain on the backend.
Core Concepts
Manage Sponsored Products, Sponsored Brands, and Sponsored Display entities through exact cataloged paths and media types.
List operations retain product-specific pagination and local ASIN/SKU filtering where the provider does not support those filters.
SP paths are Sponsored Products Advertising API paths and are unrelated to the Seller SP-API.
Operation Catalog
The concise index below is generated from the same contract consumed by the runtime. See references/api.md for full request, validation, pagination, result, status, and example details, and references/api.json for the machine-readable contract.
sb.create_ad_groups - Create Ad Groups
Request: POST /sb/v4/adGroups; access: ; execution: .
Ask for an owned connectionId, the intended account, marketplace, and region. workspaceId is optional: omit it when the API key is already bound to the workspace. Select an exact callable operation from references/api.json; do not invent paths or use a removed compatibility row. Run the relevant script from: scripts/amazon_api.py.
Reads may follow a continuation token only when fetchAll is true and always stop at maxPages. Paginated reads expose stable execution metadata: success, pagesFetched, truncated, and total when the list field is known. Client-side ASIN/SKU filters additionally report serverTotalBeforeClientFilter and clientSideFilters.
Ads entity mutations require preview, exact user approval, and a separate one-use confirmation. In the aggregate package, authorization remains a dedicated connection-lifecycle action rather than an entity mutation.
Ads requests use an owned Ads connection and the allowlisted Sponsored Products, Sponsored Brands, Sponsored Display, or reporting families. Report downloads use opaque backend tokens and optional safe outputFile publication.
Authentication and Connection Lifecycle
Set NEXSCOPE_PROXY_BASE to the approved gateway base and NEXSCOPE_API_KEY to the caller credential.
Use the workspace resolved from the API key (or an explicit workspaceId override) and an owned connectionId to select the account. Do not ask for Ads access tokens, refresh tokens, OAuth client credentials, or signing secrets.
If authorization is missing or expired, use the corresponding connection script to authorize or inspect status. Token persistence and refresh remain backend-owned.
Billing or credit behavior is determined by the gateway contract; do not repeat legacy credit claims or onboarding instructions.
python scripts/amazon_api.py '{"operation":"sb.create_ad_groups","workspaceId":"user:42","connectionId":7,"payload":{"adGroups":[{"state":"PAUSED","adGroupId":"2001","campaignId":"1001","name":"Example ad group","defaultBid":0.5}]},"phase":"preview"}'
Examples use placeholders. An Ads entity write preview is not approval; confirmation must be a separate command after the user accepts the exact preview.
Display Rules
Identify the selected account, marketplace, operation, and whether it is read-only or a mutation.
For a mutation, show the preview and wait. Never print or reuse a confirmation token outside the separate approved confirm call.
For pagination, show pagesFetched, total when known, truncated, and the continuation token only when another bounded call is required.
For client-side filters, show the provider count before filtering and the filters applied.
For a pending report or feed, preserve the returned resource ID and show the structured resume instruction.
For documents, show outputFile, byte count, content type, and decode/compression metadata when returned; never expose a provider download location.
Treat 401/403 as authentication or ownership failures, 429 as a bounded-backoff condition, and 5xx as a sanitized upstream failure. Ads report 425 recovery applies only to the Ads report workflow.
Important Limitations
ASIN and SKU filters may run client-side after the provider page is received.
Writes always require preview and a separate exact confirmation.
Provider rate limits vary by operation and account. Honor returned rate-limit metadata and Retry-After; do not hardcode a universal requests-per-second value.
The gateway allowlists exact API families. An unsupported path is a contract error, not a reason to bypass the gateway.
Official API names and versions are recorded in references/api.md; this package intentionally contains no direct external host URL.
User Expression and Scenario Quick Reference
Classification
Guidance
Applicable
Read or change SP, SB, or SD campaign entities
Not applicable
Seller Central operations or report configuration discovery
When the request crosses package boundaries, use the aggregate package only if its catalog contains the exact operation; otherwise choose the narrower package.
Privacy and Errors
Do not request, print, cache, or store provider credentials. Treat advertising account, campaign, targeting, metric, and report data as private. Scripts exit nonzero for HTTP, application, upstream, invalid JSON, and malformed response failures. See references/testing.md for executable prompts, expected routes, error cases, and evidence.
