Add ScanSci PDF integration guide

Author: Vambrocop

README.md

@@ -61,6 +61,7 @@ See:
 - [`docs/reading-list.md`](docs/reading-list.md)
 - [`docs/gu-baojing-method-extraction.md`](docs/gu-baojing-method-extraction.md)
 - [`docs/knowledge-graph-navigation.md`](docs/knowledge-graph-navigation.md)
+- [`docs/scansci-pdf-setup-and-workflow.md`](docs/scansci-pdf-setup-and-workflow.md)
 - [`docs/version-roadmap.md`](docs/version-roadmap.md)
 
 ## Origin and Workflow Inspiration

docs/scansci-pdf-setup-and-workflow.md

@@ -0,0 +1,186 @@
+# ScanSci PDF Setup and Workflow
+
+`scansci-pdf` is best treated as a literature-ingestion layer for EvidenceForge.
+
+It is not a meta-analysis method by itself. It is the infrastructure that helps an agent:
+
+- search papers;
+- resolve DOI and citation metadata;
+- download PDFs;
+- import `.bib` lists;
+- hand a structured paper set to screening, extraction, review, or meta-analysis workflows.
+
+## Why It Fits EvidenceForge
+
+EvidenceForge already covers:
+
+- protocol design;
+- search and screening logs;
+- coding sheets;
+- effect-size extraction;
+- meta-analysis and umbrella-review logic;
+- environmental and life-science review variants.
+
+`scansci-pdf` fits before those steps. It can serve as the retrieval and PDF-collection layer for:
+
+- systematic reviews;
+- meta-analysis projects;
+- umbrella reviews;
+- AI-assisted screening;
+- domain-specific evidence audits.
+
+## Recommended Architecture
+
+Use three layers:
+
+1. Retrieval layer:
+   `scansci-pdf`
+
+2. Evidence-method layer:
+   `EvidenceForge`
+
+3. Analysis or writing layer:
+   `meta-analysis-forge`, `umbrella-review-skeptic`, `environment-life-review-forge`, or downstream manuscript tools.
+
+This keeps paper retrieval separate from methodological judgment.
+
+## Local Installation
+
+Install from PyPI:
+
+```bash
+py -m pip install scansci-pdf
+```
+
+Check the local environment:
+
+```bash
+scansci-pdf check
+```
+
+The project README describes the standard MCP stdio configuration as:
+
+```json
+{
+  "mcpServers": {
+    "scansci-pdf": {
+      "command": "scansci-pdf",
+      "args": ["run"]
+    }
+  }
+}
+```
+
+For Claude Code, the command-line equivalent is:
+
+```bash
+claude mcp add scansci-pdf -- scansci-pdf run
+```
+
+Then verify:
+
+```bash
+claude mcp list
+```
+
+## Claude Code and Other Agents
+
+If an MCP-compatible agent exposes the configured server, `scansci-pdf` can provide tools such as:
+
+- keyword search;
+- DOI or arXiv download;
+- batch download;
+- citation export;
+- `.bib` import;
+- health checks and network diagnostics.
+
+That makes it a strong front end for AI-assisted evidence synthesis.
+
+## How To Use It With EvidenceForge
+
+### Example: systematic review intake
+
+1. Search papers by keyword.
+2. Export or collect DOI and citation metadata.
+3. Download PDFs in batch.
+4. Build a screening set.
+5. Move to:
+   - `evidence-synthesis-forge` for protocol and screening;
+   - `meta-analysis-forge` for coding and pooling;
+   - `environment-life-review-forge` for domain-specific extraction.
+
+### Example prompt pattern
+
+```text
+Use scansci-pdf to search and download candidate papers on rice biochar and greenhouse gas emissions.
+Then switch to EvidenceForge screening and extraction workflows.
+Keep retrieval separate from methodological judgment.
+```
+
+## Suggested Workflow Boundaries
+
+Use `scansci-pdf` for:
+
+- paper search;
+- PDF download;
+- DOI resolution;
+- citation export;
+- `.bib` import;
+- retrieval diagnostics.
+
+Do not use `scansci-pdf` as a substitute for:
+
+- study eligibility judgment;
+- risk-of-bias appraisal;
+- effect-size extraction quality control;
+- causal interpretation;
+- synthesis-model choice.
+
+## Compliance and Risk Boundary
+
+The project README lists multiple retrieval routes, including open-access sources, publisher links, WebVPN or CARSI access, and also routes such as Sci-Hub or LibGen.
+
+Because of that, EvidenceForge should document a simple boundary:
+
+- prefer open-access, publisher-authorized, institution-authorized, or otherwise legitimate access routes first;
+- treat legally or institutionally sensitive routes as a user-side decision, not an automatic default;
+- document the chosen retrieval strategy in the review ledger when reproducibility matters.
+
+The project also documents a `legal_only` strategy. That is the cleanest default when a review needs a conservative retrieval policy.
+
+## Useful Configuration Ideas
+
+Examples from the project README:
+
+- `download_strategy = legal_only` for conservative retrieval;
+- `output_dir` to keep a dedicated paper folder per review;
+- `batch_workers` for larger download jobs;
+- `vpnsci_enabled` or `carsi_enabled` when institution-authorized access is available.
+
+## What To Store In Your Own Repositories
+
+Keep these in your GitHub repositories:
+
+- setup instructions;
+- MCP registration instructions;
+- recommended retrieval policies;
+- folder conventions for downloaded PDFs;
+- review-ledger notes explaining whether retrieval was OA-only, institution-assisted, or mixed.
+
+Do not vendor the whole upstream project into EvidenceForge unless you have a specific maintenance reason. A thin integration document is cleaner and easier to keep current.
+
+## Minimal Review Ledger Add-On
+
+When `scansci-pdf` is used in a review project, record:
+
+- search date;
+- search terms;
+- download strategy;
+- whether `.bib` import was used;
+- output folder;
+- access mode: OA-only, publisher-authorized, institution-authorized, or mixed;
+- unresolved missing PDFs.
+
+## Source
+
+- Upstream repository: https://github.com/Rimagination/scansci-pdf