Turn a reading list into source-grounded notes, check a draft against them, and build a paper.
- Reading list to notes, with every quote checked against its page in the source.
- A draft checked against those notes: uncited and unverified claims are flagged.
- A paper built as PDF, HTML or DOCX, after a bibliography check.
No API key, no server: it runs inside the assistant you already use (Claude Code, or Codex through an adapter). Early-stage, version 0.2.2.
Formerly research-kit. Upgrading? Remove the old plugin first; see INSTALL.md.
Catalog · Install guide · Workflow · Lecture quickstart · What is verified · Changelog
In Claude Code (Git must be installed to download the plugin this way; the ZIP route needs none):
/plugin marketplace add nik110403/provenote
/plugin install provenote@provenote
Reading or loading a skill does not execute any tool by itself; tools run only when the assistant calls them.
Check your setup: run doctor (how). Missing pdftotext or pandoc?
Run scripts/install-deps.sh (macOS/Linux) or scripts/install-deps.ps1 (Windows);
details in INSTALL.md. Then ask Claude to start a research project,
search the literature, or verify notes.
Other routes: Codex adapter, including a ZIP and file-manager route that needs no Git, Python or terminal · create your own skill · check quotes in a saved Word lecture without Zotero or notes.
What's inside: 8 skills, 6 roles, 11 tools
| Kind | Name | What it does |
|---|---|---|
| Skill | new-research-project | Scaffold a numbered project tree from a question |
| Skill | literature-search | Search providers, save a reading list with a receipt |
| Skill | pdf-to-markdown | Convert PDFs to page-marked Markdown |
| Skill | literature-notes | Write a structured note with page-pinned quotes |
| Skill | review-against-notes | Review a draft against the notes |
| Skill | paper-build | Check the bibliography and build PDF or HTML |
| Skill | analysis-pipeline | Set up and run a repeatable analysis |
| Skill | tribunal | Second opinion on a costly decision: sealed multi-lens review with a verbatim record |
| Agent | source-reader | Read one source into a note |
| Agent | claim-checker | Judge whether notes support cited claims |
| Agent | skeptic | Challenge the strongest claims of a draft |
| Agent | tribune | One seat of a tribunal: review through a single lens |
| Agent | refuter | Try to disprove one finding |
| Agent | pipeline-runner | Run a trusted analysis pipeline and report |
| Tool | verify-quotes | Match each quote, on its stated page, against a markdown copy of the source (hash enforced with --strict) |
| Tool | check-docx | Inventory and page-verify typed Word quotations without Zotero; see Word workflow |
| Tool | lit-search | Literature search with provider receipts |
| Tool | pdf2md | PDF to Markdown with page markers |
| Tool | check-claims | Check each cited note (runs verify-quotes on it) and flag uncited or unverified claims; does not judge meaning |
| Tool | check-bib | Check bibliography structure and citation coverage |
| Tool | build-paper | Build PDF, HTML or DOCX with pandoc; Word export is one-way output |
| Tool | new-project | Scaffold a research project |
| Tool | new-pipeline | Scaffold an R or Python pipeline |
| Tool | tribunal-scribe | Hash-chained verbatim record for a tribunal workspace |
| Tool | doctor | Read-only setup diagnostics with per-OS fixes; see doctor |
The complete catalog lists inputs, outputs and dependencies; the adapter guide covers manual use with other assistants.
The demo above is a real Claude Code run on a small demo project (how it was made). The offline quote check on the bundled example below needs no PDF software and no API key. To reproduce it, clone the repository and, with Python 3.9+, run from its root:
cd examples/basic
python3 ../../plugins/provenote/bin/verify-quotes --strict 01_notes/lit/*.md
python3 ../../plugins/provenote/bin/verify-quotes --strict 01_notes/negative/*.mdThe first prints 14 of 14 PASS; the second reports the planted fabrication as FAIL (expected). See the walkthrough. This checks converted text against quotes, not the original PDF and not whether a claim follows from a quote; see what is verified for the evidence limits.
Enter at any stage for which you have the inputs. The tools do the mechanical checks; the assistant roles return text for the main session to save. See docs/workflow.md.
flowchart LR
A[Scaffold] --> B[Search]
B --> C[Acquire PDFs]
C --> D[Convert]
D --> E[Read and verify quotes]
E --> F[Draft]
F --> G[Review against notes]
G --> H[Bibliography check and build]
A -.-> I[Analysis side track]
I -.-> F
Python 3.9+ (standard library only) runs the helpers; PDF conversion needs pdftotext,
paper builds need pandoc, and an optional mailto setting gives search a contact address. Dependencies,
Windows invocation and limits: INSTALL.md.
CONTRIBUTING.md covers checks, privacy and licensing.
Report vulnerabilities privately; see SECURITY.md. What leaves your machine, and when: privacy policy.
Code, skills, agents and templates are MIT. Documentation under
docs/ is CC BY 4.0; bundled papers and conversions retain their
authors' attribution. See the licensing map.