Literature Metadata Curator
Purpose
Query, correct, and complete bibliographic metadata for a selected Zotero parent item. The workflow handles cases such as inconsistent title casing, missing authors, incomplete journal/volume/page fields, incomplete DOI/ISBN entries, and incorrectly set item types.
Inputs
| Parameter | Required | Description |
|---|---|---|
| Skip identifier fast path | No (default: off) | Bypass Zotero identifier lookup and run literature-metadata-search directly. |
Select exactly one parent item in the Zotero item list. Attachments and multiple items are not accepted.
Behavior
The workflow runs fully automatically without user confirmation. It uses these routes:
- Local fast path (default): If the item has a DOI, ISBN, or a URL that deterministically resolves to a DOI, arXiv, or PubMed identifier, the workflow calls
runtime.hostApi.metadata.translateIdentifier(a controlled read-only ZoteroTranslate.Searchfacade). When the candidate identifier matches and contains valuable bibliographic information, results are written back directly. - Forced Agent search: When Skip identifier fast path is enabled, the workflow bypasses local lookup and runs
literature-metadata-searchdirectly. The identifier remains available to the Agent as search context. - Skill-Runner fallback: When the option is off but no reliable identifier exists, local search returns no results, the translator fails, the candidate is untrusted, or the identifier does not match, the workflow runs the same skill for web-based metadata retrieval.
All routes share the same canonical result format and apply handler.
If an item already has a DOI, ISBN, or another supported identifier but the default run returns incorrect metadata, enable Skip identifier fast path to force Agent-based web search. This does not relax candidate matching or evidence requirements.
Write-back Rules
The workflow updates the parent item's bibliographic fields:
- Title, DOI, ISBN, ISSN, URL, abstract, date, language, library catalog
- Journal/conference/book/thesis/report fields (journal name, volume/issue/pages, publisher, conference name, school, report type, etc.)
- Creators (authors, institutional authors, etc.)
itemTypewhen supported by high-confidence evidence (e.g., journal article corrected to thesis)
For a paper originally published in Chinese, the Agent writes Chinese-character author names only when an authoritative source verifies the complete author list. It does not replace creators with pinyin, translated names, or guessed Chinese characters; when the complete Chinese list cannot be verified, existing creators are preserved.
The workflow does not modify attachments, notes, tags, collections, related items, PDF files, or web snapshots.
Without a stable identifier, the workflow only overwrites an existing title or changes the item type when: the candidate can be proven to be the same direct work, at least two independent bibliographic signals agree, and an authoritative landing page corroborates. Container titles are written to the appropriate container field rather than replacing the work title. Low-confidence, conflicting-candidate, or疑似-only results are skipped.
Output And Apply
Metadata changes are applied directly to the selected Zotero parent item. No intermediate confirmation step is required.
Model Recommendation
- Fast path hit (supported identifier present and the option is off): No backend model needed.
- Skip option enabled or fallback to
literature-metadata-search: A model with web search capability is recommended. The task is lightweight retrieval and evidence verification — it does not require long-form writing ability, but must distinguish homonyms, preprint vs. published versions, papers vs. theses, and different editions.
Dependencies
- Backend: Skill-Runner (for fallback after local search miss)
- Skill:
literature-metadata-search - Zotero Host API:
metadata.translateIdentifier(controlled read-only fast path) - Apply Handler:
handlers.parent.updateMetadata
Related Workflows
- Literature Search & Ingest — Search for new literature and ingest into Zotero
- Literature Analysis — Generate digest and citation analysis from PDF/Markdown
- Tag Regulator — Normalize tags after metadata is complete