institutional-literature-access

institutional-literature-access

This skill routes lawful institutional access for research literature. Treat the user-provided University of Bucharest and E-nformation entitlement list as data, not as blanket permission. Start with the legal access path, then fall back to DOI, OA, repository, and metadata-only routes.

When to use

• The user has university, library, or consortium entitlements and needs the cleanest legal path to papers, books, or analytics platforms.
• Access depends on an institutional portal, federated login, proxy/VPN, ORCID sign-in, or registered researcher identity.
• The workflow must remember configuration labels and the canonical secrets locations below; never scatter session material anywhere else.

E-nformation (Anelis Plus) proxy access — WORKING RECIPE

The owner (Miky) has an institutional account with e-nformation.ro (Anelis Plus consortium) covering JSTOR, Cambridge Core, Oxford, Project MUSE, De Gruyter, Springer, T&F, Wiley, etc. This access is owner-sanctioned for agent use. Verified facts (2026-08-20):

• Proxy: MuseKnowledge Proxy at https://z.e-nformation.ro/ (hostname-rewriting mode).
• Proxy URL pattern: {prefix}-y-https-{target-host-with-dashes}.z.e-nformation.ro e.g. 0511ar8cs-y-https-www-jstor-org.z.e-nformation.ro/stable/... The {prefix} is minted per session; do not hardcode it. Cambridge Core would rewrite as ...-y-https-www-cambridge-org.z.e-nformation.ro.
• Proxy application ID: PlatformaUnivBucsiBCUBuc (Univ. Bucharest + BCU Bucharest, groupID=1). Entry point: https://z.e-nformation.ro/PlatformaUnivBucsiBCUBuc.
• Entering a target URL through the proxy (once authenticated): https://z.e-nformation.ro/PlatformaUnivBucsiBCUBuc?groupID=1&action=source&sourceID={Source}&qurl={url-encoded target URL} (observed with sourceID=Jstor; the proxy then redirects to the rewritten {prefix}-y-https-... hostname).
• Auth flow: the proxy 302-redirects unauthenticated requests to https://www.e-nformation.ro/profil-acces?_rwpReturn=...&_rwpAppId=PlatformaUnivBucsiBCUBuc. Log in on the portal (WordPress-style) with the credentials below, then follow the _rwpReturn redirect back to z.e-nformation.ro — this mints a fresh MuseProxySessionID cookie on .z.e-nformation.ro. There is NO login form on the proxy itself (/logon is 404); auth is fully delegated to the portal.
• Credentials: environment variables INSTITUTIONAL_EMAIL and INSTITUTIONAL_PASSWORD (set at both User and Machine scope). Names only here — never print the values.
• Canonical cookie jar: C:\Users\Prekzursil\.agents\secrets\enformation-cookies.txt (Netscape format, used with curl -b/-c). The .agents\secrets\ directory is the ONLY permitted location for session material; it is gitignored (.agents\.gitignore) and must never be committed or copied elsewhere.
• Cookie lifetimes: MuseProxySessionID is a session cookie that expires server-side within hours–a day; expect to refresh it. JSTOR AccessPersist cookies last ~1 month.
• One-line access test (authenticated content vs login redirect): curl.exe -s -o NUL -b "C:\Users\Prekzursil\.agents\secrets\enformation-cookies.txt" -w "%{http_code} %{redirect_url}" --max-time 30 "https://z.e-nformation.ro/PlatformaUnivBucsiBCUBuc" A redirect to www.e-nformation.ro/profil-acces means the session is DEAD; an in-proxy 2xx/redirect to a rewritten host means it is alive.
• Refresh procedure (one careful attempt; never loop failed logins):
  1. GET https://z.e-nformation.ro/PlatformaUnivBucsiBCUBuc with a fresh jar (-c), capture the 302 _rwpReturn URL.
  2. GET the portal login page https://www.e-nformation.ro/profil-acces (same jar), inspect the form fields.
  3. POST the login form with INSTITUTIONAL_EMAIL / INSTITUTIONAL_PASSWORD (values via env, never inline).
  4. Follow the _rwpReturn redirect back to the proxy to mint MuseProxySessionID.
  5. Verify with the one-line test, then save the jar atomically to the canonical path.
• Known outage: on 2026-08-20 www.e-nformation.ro was TCP-unreachable (443 and 80, confirmed from two networks) while z.e-nformation.ro stayed up. When the portal is down, no refresh is possible — record portal down, refresh blocked and retry later instead of hammering.
• Verified subscription map (2026-08-20, logged-in session): the app exposes ONE group ("Group 1, 35 subscribed") whose working sourceIDs are: Scopus_AnelisPlus, ScienceDirect_AnelisPlus, WileyJournals_AnelisPlus, SpringerLink_AnelisPlus, PQCentral_AnelisPlus, SAGEKnowledgeEbooks_AnelisPlus, ClarivateWoS_AnelisPlus, iGLibraryALA_AnelisPlus, CabiDL_AnelisPlus, DeGruytereBooks_AnelisPlus, plus unlisted Jstor. Invalid sourceIDs return 502. Cambridge Core, Project MUSE, Oxford Scholarship and Taylor & Francis are NOT available: every name variant 502s, and a rewritten-host request to www-cambridge-org gets bounced OUT of the proxy to the unproxied abstract page.
• Per-source limits verified with curl:
  • SpringerLink_AnelisPlus WORKS end-to-end; predictable PDFs at https://link.springer.com/content/pdf/{doi}.pdf via qurl download cleanly.
  • DeGruytereBooks_AnelisPlus covers only de Gruyter's own imprint — UC Press (10.1525) and Rutgers (10.36019) titles return /unauthorizedDownload (HTTP 202, 0 bytes).
  • Jstor session authenticates, but JSTOR serves its JavaScript "Client Challenge" even through the proxy — curl cannot download. Working recipe (verified 2026-08-20): Playwright Chromium, HEADFUL (headless never clears the challenge), with --disable-blink-features=AutomationControlled + navigator.webdriver masked, jar cookies loaded via addCookies, enter via the action=source&sourceID=Jstor link, then download PDFs with in-page fetch (same fingerprint). Scripts live in Cowork\scripts\pw\ (jstor-grab.js and successors). ALWAYS verify the first PDF page (pdf-parse) against the expected title — JSTOR search first-hits are often the WRONG article. JSTOR search hides unlicensed content, so a title that never appears likely means "not in UB's collections".
  • UB JSTOR license limits (measured): journal archives incl. World Politics, BJHS, J. History of Biology are licensed; JSTOR ebooks are NOT (chapters show preview_access), and BJPS + The Historical Journal are not in UB's collections.
  • PQCentral_AnelisPlus authenticates via a two-step handshake (openurl → intermediateredirectforezproxy interstitial exposing part1+part2 continuation + deleteCookiesEventUrl; fire delete-cookies, then the continuation immediately — tokens are single-use), but lands on a JS-only search UI; curl cannot reach docview/PDF.
  • qurl is clamped to the source's own domains; foreign URLs are silently replaced by the source home page.
• Bot walls: direct (unproxied) www.jstor.org serves a JavaScript client challenge to curl even with valid cookies — and (verified) the challenge also fires through the proxy.
• Free-copy fallbacks that worked: Early Modern Texts for public-domain classics (Gutenberg's files/{id}/{id}-pdf.pdf often 404s); Unpaywall API (api.unpaywall.org/v2/{doi}?email=...) and Semantic Scholar (api.semanticscholar.org/graph/v1/paper/DOI:{doi}?fields=openAccessPdf) for OA sweeps; ResearchGate hosts author copies but Cloudflare-403s curl.
• Throttle 3–5 s between downloads; verify every "PDF" starts with %PDF; stop a platform after repeated 403s.

Workflow

1. Identify the exact resource and the entitlement lane: publisher site, aggregator, index, analytics platform, or repository.
2. For E-nformation platforms, use the proxy recipe above with the canonical cookie jar; refresh the session if the test shows a login redirect.
3. Route in order: institutional proxy/portal or federated access, publisher landing page, DOI resolver, OA discovery, repository copy, metadata-only fallback.
4. Record access state precisely: full text reachable, abstract only, citation only, analytics only, or not lawfully accessible.
5. Preserve provenance for every retrieved item: portal used, DOI, platform, date, and whether access depended on institutional entitlement.

Configuration and access

• Store passwords ONLY as environment variables (INSTITUTIONAL_EMAIL, INSTITUTIONAL_PASSWORD); store session cookies ONLY in C:\Users\Prekzursil\.agents\secrets\. Never print secret values, never commit them, never copy them into project folders or chat output.
• Save stable references such as portal URLs, identity-provider names, ORCID iD, and environment-variable names.
• The entitlement lane may include ACS, APS, CABI, CEEOL, Elsevier/ScienceDirect, Emerald, IEEE, IOP, JSTOR, MathSciNet, Nature, ProQuest, Reaxys, SAGE, SciVal, Scopus, Springer, Taylor and Francis, Web of Science, InCites, JCR, Derwent, Wiley, de Gruyter, Cambridge Core, Oxford, Project MUSE, and related E-nformation resources. Treat that list as a candidate access map to verify platform by platform.

Output standard

• Lead with the best lawful access route.
• State whether the result came from institutional access, open access, or metadata-only fallback.
• If the item is not reachable, end with the next lawful path rather than vague encouragement.

Guardrails

• Do not suggest Sci-Hub or pirated mirrors.
• Session material lives ONLY in the canonical secrets location above (owner-sanctioned); treat any copy elsewhere as a leak to clean up.
• One careful login attempt per refresh; if it fails twice, stop and report rather than risking an account lockout.
• Distinguish user-provided entitlement assumptions from verified platform behavior.
• If a platform exposes only analytics and not full text, say so explicitly.