git:20260906.3d137c0 to git:20261001.2d71fc1

15 added, 7 removed. Audit A to A.

---

name: scrapingdog

- description: Collect live public web data through ScrapingDog when an API key is available. Use for search, scraping and dedicated public-data endpoints.

+ description: Collect specialized public data through ScrapingDog when its configured route is selected. Use for Maps, social, commerce, jobs and other dedicated endpoints; generic web discovery and public-page acquisition belong to research.

---

# ScrapingDog

Resolve this skill directory before running its `scripts/key-env-check.sh`.

The checker reports whether the key is available in the current or interactive

shell without printing it. Never expose a key, raw account response or

credential-bearing request URL.

+ Use `research` and its configured Scrapinho MCP for generic web discovery and

+ public-page acquisition. This package owns specialized endpoints without

+ verified parity, not an automatic fallback from a refused request.

Inspect the available MCP tools and select the dedicated endpoint for the task.

- Use an existing HTTP helper when the endpoint is not exposed by MCP. Read only

- the relevant endpoint family below. Generic scraping is for pages that have no

- more appropriate dedicated endpoint; send the rendering mode explicitly.

+ Use an existing HTTP helper when that specialized endpoint is not exposed by MCP.

+ Read only the relevant endpoint family below; send rendering mode explicitly.

For paid batches, use `scripts/account_summary.sh` before and after collection.

- Respect an explicit budget. Bound retries and concurrency, cache useful results,

- and record failures. After a missing key or bounded provider failure, disclose

- the reason before using Firecrawl or host search. Do not submit private or

+ Respect the user's selected scope and budget; a low price is not authorization.

+ Bound retries and concurrency, cache useful results, and record failures.

+ After a missing key or non-refusal provider failure, disclose the reason before

+ using an available specialized alternative. Stop on CAPTCHA, 403 or 429; do not

+ change tools, providers or proxies to bypass refusal. Do not submit private or

session-authenticated content to a public scraping provider.

+ Prompt-based `/chatgpt` and other external inference modes are not substitutes

+ for the configured subscription. Invoke them only when the user explicitly

+ authorizes that service and its costs; a configured key does not grant permission.

- [endpoint-index.md](references/endpoint-index.md): choose an endpoint when the

family is unclear; historical costs are not current price evidence.

- [core-tools.md](references/core-tools.md): static/dynamic pages, account and proxy.

- [google-serp.md](references/google-serp.md): search, news, trends and search answers.

- [local-maps.md](references/local-maps.md): places, reviews and local businesses.

- [social-video.md](references/social-video.md): public social data and transcripts.

- [commerce-travel.md](references/commerce-travel.md): products, travel and marketplaces.

- [b2b-research.md](references/b2b-research.md): jobs, company data and Scholar.

Use parameter objects or URL encoding for HTTP queries, a bounded timeout and

sanitized errors. Do not print arbitrary error bodies that may echo credentials.

Direct known official sources and local evidence need no proxy when they already

provide the requested information.