← Về thư mục
📄 / / root / .hermes / skills / research / academic-source-retrieval / SKILL.md

name: academic-source-retrieval description: "Skill for finding, retrieving, and downloading full-text academic papers using DOIs, URLs, or metadata." version: 1.0.0 author: Hermes Agent license: MIT platforms: [linux] metadata: hermes: tags: [Research, PDF, Download, Academic, Paper, DOI, Sci-Hub, OpenAccess] related_skills: [api-credential-acquisition]

Academic Source Retrieval

Skill for finding, retrieving, and downloading full-text academic papers using DOIs, URLs, or metadata.

Triggers

Procedure for Challenging Publishers (e.g. T&F)

  1. Avoid direct navigation: When encountering T&F or publishers with aggressive Cloudflare (403/Turnstile), do NOT repeatedly navigate; it blacklists the VPS IP.
  2. DOI-Centric Retrieval: Always prioritize DOIs via academic-source-retrieval fallback chain (Unpaywall -> Core -> SciHub) rather than browser scraping.
  3. Workspace Discipline:
    • EVERY task that involves downloading/finding a paper must be executed within a clear workspace folder (/opt/ai-os/products/ceo/content/research/<topic>/).
    • After claim task, worker MUST immediately mkdir -p <workspace_path>/downloads and put retrieved papers there.
    • RULE: If the task involves retrieving academic materials, the worker MUST assign the workspace_path to the task metadata upon claiming. Never leave it empty.
  4. Verification: Report the final URL to the user using the Tailscale Funnel file link format (e.g., https://vmi3427693.tail8c1aaf.ts.net/files/opt/ai-os/products/ceo/content/research/<path>). Never send raw local paths.

Pitfalls & Workarounds

🛡️ CAPTCHA & Cloudflare

Many mirrors and publishers (Taylor & Francis, Sci-Hub) use Cloudflare Turnstile or Altcha. - Symptom: browser_navigate shows "Just a moment..." or "Verification". - Limitation: Agent cannot solve these captchas directly. If encountered, report the specific verification type to the user.

Search Reconstruction Strategy for Institutional Profiles

If standard web extraction fails (e.g., due to CAPTCHAs, 403 errors, or timeouts on university websites), do not persist in direct navigation. Immediately pivot to "Search Reconstruction": 1. Targeted Queries: Use site-specific queries combined with keywords like Institutional Research, Institutional Effectiveness, director, analyst, job description, or qualifications. 2. Snippet Synthesis: Read the description snippets from multiple search results (LinkedIn, university career pages, news articles, PDFs) to piece together the organizational hierarchy, roles, and common qualification requirements. 3. Cross-Verification: Compare terminology across the search results to infer standard industry practices (e.g., distinguishing between a 'Strategic Planning' role versus a purely 'Data Analytics' role). 4. Reporting: Explicitly state in the report that the data was reconstructed via institutional search to provide transparency about the method.

Anti-Hyphen Formatting Rule (Vietnamese Reports)

Anti-Hyphen & Factcheck Formatting Rule (Vietnamese Reports)

When reviewing, authoring, or fact-checking reports for Vietnamese projects under research-critism rules or when specifically requested: - No Bullet Hyphens: Do NOT use hyphens (-) for list items or bullet points in prose and lists. - Replacement: Replace all item hyphens with asterisks (*) or numbered lists. - No Hyphens in Prose Sentences: Do not use hyphens as dashes/connectors within sentences or lists (except when forming part of a hyphenated word or compound). - In-text Citations: Every claim must be backed by an in-text citation (e.g., (Harvard University, n.d.)) referencing the sources or local files. - Reference Local Files: List local offline source files directly (e.g., nd_ospir_bailey.html) instead of masking them behind external/Tailscale URLs. - Factcheck / Citation Verification Protocol: 1. Identify exact dates, project names, and individual names (e.g., matching cựu học viên/chức vụ/chương trình). 2. Verify historical status at the time of the event (e.g., whether a person was Director of Planning or Director of Finance at that specific date). 3. Detail changes in organizational/government structures (like merged departments) to provide clear context on dates/roles. 4. Note discrepancies in titles (e.g. distinguishing a "Student Researcher / PhD student" from a "Doctor/PhD" at the event date).

🔑 API Keys

paper-search-mcp functions best with API keys for CORE, DOAJ, and Semantic Scholar. Without them, you will hit rate limits (HTTP 429) quickly.

Verification