Stats
Actions
Tags
From weekly-lit-review
Run the genomics literature review pipeline. Searches bioRxiv and 15 journal RSS feeds, downloads PDFs, then Claude reads each PDF and writes critical reviews with scores. Supports an interactive daily mode (--interactive): query titles/abstracts, pick papers from a selectable list, then deep-review only the chosen ones. No API key needed — works entirely within Claude Code. Triggers: "literature review", "weekly papers", "journal scan", "paper review", "genomics review", "preprint screening", "lit review", "daily papers".
How this skill is triggered — by the user, by Claude, or both
Slash command
/weekly-lit-review:weekly-lit-review [--interactive] [--days N] [--max-papers N] [--no-pdf] [--doi DOI [--doi DOI ...]][--interactive] [--days N] [--max-papers N] [--no-pdf] [--doi DOI [--doi DOI ...]]The summary Claude sees in its skill listing — used to decide when to auto-load this skill
A literature-review pipeline. To keep context small, the detailed steps live in per-mode
A literature-review pipeline. To keep context small, the detailed steps live in per-mode reference files — read only the one that matches the invocation, plus the shared review/score procedure.
Pick exactly one mode from $ARGUMENTS, then read that file and follow it:
| If arguments contain… | Read and follow |
|---|---|
--interactive | ${CLAUDE_PLUGIN_ROOT}/skills/weekly-lit-review/references/mode-interactive.md |
--gmail-stork | ${CLAUDE_PLUGIN_ROOT}/skills/weekly-lit-review/references/mode-gmail-stork.md |
one or more --doi | ${CLAUDE_PLUGIN_ROOT}/skills/weekly-lit-review/references/mode-doi.md |
| otherwise (default) | ${CLAUDE_PLUGIN_ROOT}/skills/weekly-lit-review/references/mode-batch.md |
Every mode reuses the shared procedure in
${CLAUDE_PLUGIN_ROOT}/skills/weekly-lit-review/references/review-and-score.md
(how to read a paper, the 7 review sections, 0–10 scoring, and how reviews/summary HTML are
produced via the render scripts). Read it when you reach the review step.
Ensure directories exist:
mkdir -p ~/Desktop/Claude/week-lit-review-results/{source,reviews,$(date +%Y-%m-%d)}
Layout: source/ (shared downloaded files), reviews/ (shared review HTML + JSON
archive — the canonical store for dedup), {YYYY-MM-DD}/ (per-run manifest, logs,
_index.json, and summary.html).
Read the config for search parameters (biorxiv_categories, genomics_keywords,
journal_feeds, review_engine):
Read: ${CLAUDE_PLUGIN_ROOT}/assets/config.yaml
{TODAY} below means $(date +%Y-%m-%d).
scripts/render_review.py,
scripts/render_summary.py) — do not write HTML by hand. Per-paper reviews come from
compact JSON; the summary is built in pure Python from the persisted review JSONs (no
hand-written summary JSON). Details and schemas are in references/review-and-score.md.~/Desktop/Claude/week-lit-review-results/reviews/{filename}.{html,json}~/Desktop/Claude/week-lit-review-results/{TODAY}/summary.htmlCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub zhongguojie1998/week-lit-review --plugin weekly-lit-review