web-manual-tier-quick-refs

# Recovery tier index Open this after `web` browse returns typed unsupported content or `NO_TEXT_BLOCKS`. Choose one route; do not silently escalate through several services. | Need | Load | Boundary | |---|---|---| | Direct PDF or known paper identifier | [Tier 0](../tier-0-pdf.md) | Download only an authorized public URL; extract locally. | | DOI, arXiv, PMID, or free structured API | [Tier 1](../tier-1-apis.md) | Verify the current vendor API and access policy. | | Ordinary static article/blog/documentation | [Tier 1.5](../tier-1-5-trafilatura.md) | No JavaScript or login. | | Tables, lists, metadata, known selectors | [Tier 2](../tier-2-beautifulsoup.md) | Requests + BeautifulSoup; no browser interaction. | | JS-rendered public page | [Tier 3](../tier-3-playwright.md) | Verify installed API; use the documented Nature/Springer wait mode. | | Authorized external page-to-markdown service | [Tier 4](../tier-4-jina-firecrawl.md) | Check current account, quota, price, and access policy first. | | Discovery through an external search vendor | [Tier 5](../tier-5-ai-search.md) | Explicit selection only; not a built-in `web` engine. | For forms, login, uploads, SPA interaction, or human verification use [agent-native-browser](../agent-native-browser.md), not a tier or `web browse`. Before executing a retained helper or trusting an asset, read its [known limitations](../maintenance-bundles/SKILL.md#known-legacy-limitations).