home / skills / bahayonghang / my-claude-code-settings / zoterosynth

zoterosynth skill

safe

/content/skills/academic-skills/zoterosynth

This skill searches, analyzes, and synthesizes Zotero libraries via zotero-mcp to generate literature reviews and export BibTeX.

npx playbooks add skill bahayonghang/my-claude-code-settings --skill zoterosynth

Review the files below or copy the command above to add this skill to your agents.

Files (10)

SKILL.md

2.8 KB

---
name: zotero-synth
description: Search, browse, and analyze Zotero libraries via zotero-mcp to summarize papers, generate topic or collection literature reviews, synthesize evidence with backlinks, and export BibTeX. Use when tasks involve Zotero collection browsing, collection-based review writing, paper summary, or evidence synthesis.
category: knowledge-management
tags: [zotero, literature-review, academic, research, synthesis, bibtex]
---

# ZoteroSynth

## 1) Environment Gate
!`zotero-mcp version 2>/dev/null || echo '{"status":"error","message":"zotero-mcp is NOT installed!","install":"uv tool install \"git+https://github.com/54yyyu/zotero-mcp.git\" && zotero-mcp setup","hint":"Install zotero-mcp first, then retry this skill. DO NOT proceed without it."}'`
If the check returns an error, stop and reply:
```bash
zotero-mcp 未安装。请先运行：
uv tool install "git+https://github.com/54yyyu/zotero-mcp.git"
zotero-mcp setup
```

## 2) Entry Workflow

Arguments: `<task> [query-or-item]`

- `check` → verify MCP and Zotero connectivity. Follow `references/ERRORS.md` on failure.
- `summarize <query|key>` → single-paper flow. Read `references/WORKFLOWS.md` § 1, output per `assets/prompts/summarize.md`.
- `review <topic|collection_key|collection_name>` → multi-paper review (Map-Reduce). For collection-based review, follow `references/WORKFLOWS.md` § 2 and output strictly per `assets/prompts/review.md`.
- `synthesize <question>` → evidence synthesis. Read `references/WORKFLOWS.md` § 3, output per `assets/prompts/synthesize.md`.
- `extract <query|collection>` → Extract papers to `papers.json` using `scripts/extract_papers.py`.
- `bibtex <query|keys>` → export citations via `zotero-mcp:zotero_get_item_metadata` with `format="bibtex"`.

## 3) Output Rules
- **Literature Review**:
    - Use `assets/prompts/review.md` as the canonical output contract.
    - Must cover at least 50 papers (if topic specified) or 80 papers (if no topic, broad review), when available.
    - Must include: `itemType`, `year`, `itemLink` (unchanged), `dataSource`, `dataSource url`.
    - Must include both detailed per-paper analysis and a final summary table.
- Add a backlink for each claim: `[Author, Year, item_key]`.
- Mark missing evidence as `[需确认]`; do not fabricate content.
- For >10 papers, use Map-Reduce and keep per-paper traceability.

## 4) Error Handling
- Follow `references/ERRORS.md` for user-facing error responses.
- Before semantic retrieval, check index status with `zotero-mcp:zotero_get_search_database_status`.
- If full text is unavailable, fallback to metadata + annotations and state limitations.

## 5) Resource Index
- Tool matrix: `references/TOOLS.md`
- Workflow specs: `references/WORKFLOWS.md`
- Output prompts: `assets/prompts/summarize.md`, `assets/prompts/review.md`, `assets/prompts/synthesize.md`
- Advanced setup: `references/ADVANCED.md`

Overview

This skill connects to a Zotero library via zotero-mcp to search, browse, summarize, and synthesize literature. It supports single-paper summaries, collection-based literature reviews, evidence synthesis with claim backlinks, and BibTeX export. The skill enforces traceability and explicit handling of missing full text or evidence.

How this skill works

The skill calls zotero-mcp endpoints to verify connectivity, inspect collections, and fetch item metadata or full text when available. For multi-paper tasks it uses a Map-Reduce workflow to analyze many items while preserving per-paper traceability and backlinks. Outputs follow strict templates for summaries, reviews, and syntheses and flag missing evidence explicitly.

When to use it

Summarize a single paper by key or query.
Produce a collection-based literature review or topic review across many items.
Synthesize evidence across a set of papers to answer a focused research question.
Export citation metadata or BibTeX for items or collections.
Extract a batch of papers into a JSON file for downstream processing.

Best practices

Ensure zotero-mcp is installed and configured and Zotero sync is accessible before running tasks.
Provide precise collection keys or clear topic queries to target relevant papers and meet minimum coverage requirements.
For large reviews, accept the Map-Reduce flow so the skill can preserve per-paper analysis and backlinks.
State whether full text is essential; if unavailable the skill will rely on metadata and annotations and will mark limitations.
Request BibTeX export only for specific keys or narrow queries to avoid oversized outputs.

Example use cases

Generate a concise summary of a single article to prepare an annotated bibliography.
Compile a literature review for a defined topic or Zotero collection, with per-paper analysis and a final synthesis table.
Synthesize evidence on a research question and produce claim-level backlinks to supporting items.
Bulk-extract collection items into papers.json for machine processing or sharing with collaborators.
Export BibTeX entries for a subset of items to populate a manuscript reference list.

FAQ

What prerequisites are required to run this skill?

You must have zotero-mcp installed and configured with access to your Zotero library. The skill verifies connectivity before proceeding and will report errors if the tool or index is missing.

How many papers will a review cover?

The review flow targets at least 50 papers when a topic is specified and 80 papers for broad reviews, when that many relevant items exist in the library. If fewer items are available the output will note gaps and mark missing evidence.

How does the skill handle missing full text?

If full text is unavailable, the skill falls back to metadata and annotations, states the limitation explicitly, and marks claims that lack direct evidence so you can follow up.