semantic-scholar-library-feed

1---
2name: semantic-scholar-library-feed
3description: Work with a user's Semantic Scholar account to read Research Feeds, inspect private Library folders, add papers to folders, and resolve Semantic Scholar paper records from identifiers such as arXiv IDs.
4---
5 
6# Semantic Scholar Library & Feed
7 
8## Overview
9 
10Use this skill to work against Semantic Scholar's authenticated Library and Research Feed surfaces without browser-driven login flows. Prefer the bundled CLI for Cookie Store inspection, browser-curl import, SSR probing, feed pagination, folder inspection, folder writes, and Graph `paper/batch` lookups.
11 
12## Quick Start
13 
141. Check whether the fixed Cookie Store already exists:
15 
16```bash
17python3 scripts/semantic_scholar_cli.py cookie-summary
18```
19 
202. If the Cookie Store is missing or stale, ask the user to copy an authenticated Semantic Scholar request as curl from browser DevTools, then import it:
21 
22```bash
23python3 scripts/semantic_scholar_cli.py import-curl \
24  --curl-file /tmp/semantic-scholar-request.sh
25```
26 
27`import-header` is still available when the user already extracted only the raw Cookie header.
28 
293. Check cookie health:
30 
31```bash
32python3 scripts/semantic_scholar_cli.py cookie-summary
33```
34 
354. Choose the task module below.
36 
37## Cookie Store
38 
39Save Semantic Scholar auth state under:
40 
41- `~/.auth/semantic-scholar.cookies.json`
42- `~/.auth/semantic-scholar.cookie-header.txt`
43 
44Treat `sid` and `s2` as the minimum required cookies for private Library and Feed access. If either is missing, ask the user for a fresh browser-copied curl and re-import it before touching private endpoints.
45 
46Read [references/auth-and-cookies.md](references/auth-and-cookies.md) when you need the Cookie Store workflow or the curl import format.
47 
48## Task Modules
49 
50### Research Feed
51 
52Use this path when the user wants feed export, history crawl, or local coarse filtering.
53 
541. Probe SSR if you need to inspect `var DATA`:
55 
56```bash
57python3 scripts/semantic_scholar_cli.py ssr-dump --list-names
58```
59 
602. Crawl the feed through `/api/1/library/folders/recommendations`:
61 
62```bash
63python3 scripts/semantic_scholar_cli.py feed-crawl \
64  --output /tmp/research-feed.json
65```
66 
673. Persist output after every window. Do not wait for the entire crawl to finish.
68 
69Read [references/research-feed.md](references/research-feed.md) when you need the SSR decode order, the real API path, or the pagination stop rules.
70 
71### Library Folder
72 
73Use this path when the user wants folder contents, folder diffs, or bulk add operations.
74 
751. Export a folder:
76 
77```bash
78python3 scripts/semantic_scholar_cli.py folder-entries \
79  --folder-id 13895811 \
80  --all-pages \
81  --output /tmp/folder.json
82```
83 
842. Add a paper to a folder:
85 
86```bash
87python3 scripts/semantic_scholar_cli.py folder-add \
88  --paper-id 25f612200a3821c71b99819cd671f2e60df5b470 \
89  --paper-title 'AgentArk: Distilling Multi-Agent Intelligence into a Single LLM Agent' \
90  --folder-ids 13895811
91```
92 
93Read [references/library.md](references/library.md) when you need endpoint behavior, `pageSize` limits, or the verified `entries/bulk` request shape.
94 
95### Paper ID Resolution
96 
97Use Graph `paper/batch` when BibTeX already contains stable identifiers such as arXiv IDs.
98 
99```bash
100python3 scripts/semantic_scholar_cli.py graph-batch \
101  --ids ARXIV:2602.08234,ARXIV:2602.12670
102```
103 
104Prefer this over search-page scraping when possible.
105 
106## Operating Rules
107 
108- Prefer direct HTTP after cookies exist.
109- Do not use Playwright for login. If the Cookie Store is missing, ask the user to copy an authenticated browser request as curl and import it.
110- Use Playwright only when rendered-page behavior or network inspection is needed to discover hidden interfaces.
111- Treat browser clicking as a reconnaissance step, not the main extraction path.
112- For feed history, stop on `empty days`, missing `nextWindowUTC`, or repeated windows.
113- For folder sync, resolve `paperId` first, diff against existing folder entries, then call `folder-add`.
114- If the task depends on a private page and returns `401`, ask for a fresh browser-copied curl and refresh the Cookie Store before debugging the endpoint.
115