---

name: metadata-and-sources description: "Use PaperQA metadata clients and external paper/trial source helpers safely without local parsing or RAG querying." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Metadata and Sources

Use this sub-skill when the task is about PaperQA metadata hydration or external source acquisition, not about parsing local documents or answering over an existing Docs collection.

Route here for

• Looking up paper metadata with DocMetadataClient from title, doi, authors, and requested fields.
• Choosing DEFAULT_CLIENTS, ALL_CLIENTS, or a custom ordered provider set for Crossref, Semantic Scholar, OpenAlex, Unpaywall, journal quality, and retraction checks.
• Understanding how Docs.aadd(..., title=..., doi=..., authors=...) uses metadata hydration and how upgrade_doc_to_doc_details merges metadata back into a Doc.
• Enabling ClinicalTrials.gov search through PaperQA settings and the clinical_trials_search tool.
• Using OpenReview or Zotero helpers after the user has authorized network access, downloads, optional extras, and credentials.

Route away from here

• Local PDF/text/HTML/Office/image parsing, chunking, and parser extras: use the docs-and-parsing sub-skill.
• Asking questions over loaded papers, Docs.aquery, ask, agent_query, evidence, and answer generation: use the agentic-rag sub-skill.
• pqa CLI and index build/search workflows: use the cli-and-indexing sub-skill.
• Model, embedding, prompt, LiteLLM, and settings JSON details outside source-specific tool selection: use the settings-and-configuration sub-skill.

Operating sequence

1. Identify the source intent: metadata-only hydration, metadata during Docs.aadd, ClinicalTrials.gov search, OpenReview paper discovery, Zotero library ingestion, or third-party paper acquisition advice.
2. For metadata clients, read metadata-clients.md before composing providers or selecting fields.
3. For ClinicalTrials.gov, OpenReview, Zotero, and paper-acquisition boundaries, read clinical-trials-and-literature-sources.md.
4. For failures, read troubleshooting.md before retrying live services.
5. Prefer the bundled no-network scripts for local inspection and formatting checks:
   • scripts/inspect_metadata_clients.py
   • scripts/mock_clinical_trials_format.py

Safety defaults

• Do not run live metadata, ClinicalTrials.gov, OpenReview, Zotero, or paper-download calls unless the user explicitly wants network access and accepts provider terms, credentials, and rate-limit implications.
• Use DEFAULT_CLIENTS or a small custom provider list before ALL_CLIENTS for fast DOI/title hydration.
• Do not recommend third-party scraping as a default paper acquisition path. Prefer user-supplied PDFs, authorized APIs, institutional access, OpenReview, Zotero, and direct open-access URLs.