Claude Skill MartBrowse skillsWhat is a Skill?
 
← Back to list

research-lookup (Manuscript Evidence Compiler)

Turns a research question into a manuscript-ready packet of ~60 verified academic references with an evidence matrix, claim-source map, and BibTeX.

EducationAdvanced★ 33,030⑂ 3,248AI score 8/10Last updated: Aug 9, 2026

What it does

Uses Parallel Search + Extract as the default engine to compile the external evidence needed for a scientific manuscript or research brief.

  • Runs six bounded search passes: recent primary studies, systematic reviews/meta-analyses, seminal work, methods and validation literature, contradictory/null/replication findings, and an unrestricted companion pass.
  • Deduplicates candidates by DOI, PMID, canonical URL, and normalized title, then verifies authors, year, venue, design, sample size, quantitative findings, limitations, and retraction status via Extract.
  • Writes an auditable file set: packet.md, references.bib, evidence-matrix.json, claim-source-map.json, synthesis.json, coverage.json, and search-ledger.json.
  • The target is 60 verified, unique references — shortfalls are reported in coverage.json instead of padded with weak or duplicate records.

Who it's for

  • Graduate students and researchers who need large, citation-ready evidence sets for Introduction and Discussion sections.
  • Anyone who wants supporting and conflicting evidence for a claim, not just agreement.
  • Writers scoping research gaps and methods precedent before drafting.
  • For a PRISMA-style systematic review, use literature-review instead.

Examples

  1. Manuscript packet: build manuscript-context.json (question, study type, population, comparator, outcomes) and run with --academic --target-references 60 --packet-dir sources/manuscript-research to get BibTeX plus an evidence matrix in one pass.
  2. Deep report: when the user explicitly asks for exhaustive coverage, run --force-backend research --processor pro for a long-form synthesized report.
  3. Quick fact check: for the latest official guidance without 60 references, use --no-academic --search-mode basic.
  4. Batch topics: --batch "query one" "query two" produces a separate packet directory per query and isolates failures.

· · · Install guide · · ·

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/research-lookup folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/research-lookup/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/K-Dense-AI/scientific-agent-skills.git && mkdir -p ~/.claude/skills && cp -r scientific-agent-skills/skills/research-lookup ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal and clone the repo: git clone https://github.com/K-Dense-AI/scientific-agent-skills.git
  2. Copy the skill into your Claude skills folder: mkdir -p ~/.claude/skills && cp -r scientific-agent-skills/skills/research-lookup ~/.claude/skills/
  3. Install the search CLI: uv tool install "parallel-web-tools[cli]==0.7.1" (install uv first with pip install uv if needed).
  4. Authenticate: parallel-cli login (use parallel-cli login --device on headless machines). Verify with parallel-cli --version and parallel-cli auth.
  5. If you plan to use the explicit Chat backend, export PARALLEL_API_KEY from your shell profile. Never pass the key as a command-line argument or print it.
  6. (Optional) Export OPENROUTER_API_KEY only if you want the Perplexity alternative or failure fallback.
  7. Restart Claude Code and ask something like "gather literature and references for this research question". Run commands from the repo root so the example script paths resolve.
View source on GitHub ↗License: MIT