literature-helper-polsci

literature-helper-polsci

A Claude Code skill for producing curated, citation-verified literature surveys for political science and computational social science research projects.

What it does

• Retrieves, verifies, screens, and synthesizes scholarship into a traceable, hyperlinked deliverable (rough_literature_survey.md)
• Enforces a zero-fabrication policy with anti-model-memory rules — every citation must be externally verified
• Covers three required pillars: Substantive theory, Empirical comparanda, and Methodological precedents
• Produces narrative-driven surveys (not just lists) that position your project within existing scholarship

Installation

Install as a Claude Code plugin from GitHub:

# Step 1: Add the plugin from GitHub
claude plugin marketplace add LIANJie-Jason/literature-helper-polsci

# Step 2: Install it
claude plugin install literature-helper-polsci

Structure

literature-helper-polsci/
├── .claude-plugin/
│   ├── marketplace.json        # Repo-level plugin index
│   └── plugin.json             # Plugin manifest
├── skills/
│   └── literature-helper-polsci/
│       ├── SKILL.md            # Main skill (principles, workflow summary)
│       └── references/
│           └── workflow-reference.md  # Detailed 9-step workflow
├── README.md
└── LICENSE

Dependencies

This skill works best with:

• Zotero MCP — for searching your personal Zotero library (Zotero-first retrieval)
• zotero-mcp-code skill — for efficient multi-strategy Zotero searches
• Web search tools — perplexity-search, research-lookup, or deep-research MCP for external retrieval

The skill degrades gracefully if Zotero is unavailable (falls back to API + web only mode).

Workflow

Step	Action
0	Understand project (RQ, constructs, search plan)
1	Retrieve (Zotero-first, then APIs + web)
2	Citation-network expansion
3	Deduplicate
4	Screen and tag by pillar
5	Quality control (confidence grades, DOI verification)
6	Write deliverable
7	Self-check for consistency
8	Iterative validation loop

Output

The skill produces rough_literature_survey.md containing:

• Project-to-Literature Map with cluster analysis and cross-cluster synthesis
• Project Positioning identifying gaps, contributions, and scope conditions
• Annotated Bibliography (25–50 items) with hyperlinked citations and confidence grades
• Must-Read Shortlist (10–15 items)
• Coverage Report and Audit Log for full transparency

Caveats

• You should always read the papers themselves, and this tool is merely a starting point.
• Despite all the guardrails, I suggest you to debug the final reports one to two times manually on both the accurate citations and the correct summarizations.

License

MIT