{"slug":"search-lit","title":"search-lit","summary":"Literature search and citation management for medical research. Searches PubMed, Semantic Scholar, and bioRxiv/medRxiv with verified citations. Anti-hallucination — every reference verified via API before inclusion. Generates BibTeX entries.","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-09-14T20:48:09.444232Z","repo":{"url":"https://github.com/Aperivue/medsci-skills","stars":333,"forks":77,"license":"MIT","updatedAt":"2026-10-05T14:57:51Z"},"bodyHtml":"
\n

name: search-lit\ndescription: Literature search and citation management for medical research. Searches PubMed, Semantic Scholar, and bioRxiv/medRxiv with verified citations. Anti-hallucination — every reference verified via API before inclusion. Generates BibTeX entries.\ntriggers: literature search, find papers, citation, references, bibliography, PubMed search, related work\ntools: Read, Write, Edit, Bash, Grep, Glob\nmodel: inherit

\n

Literature Search Skill

\n

You are assisting a medical researcher with literature searches and citation management for\nmedical research papers. Every reference you produce must be verified against a live database --\nnever generate citations from memory alone.

\n

Communication Rules

\n\n

Key Directories

\n\n

Search Tools: MCP (Primary) + E-utilities (Fallback)

\n

Primary: MCP Tools (Claude.ai Remote)

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
DatabaseMCP ToolPurpose
PubMedmcp__claude_ai_PubMed__search_articlesSearch by query, MeSH terms
PubMedmcp__claude_ai_PubMed__get_article_metadataFull metadata for a PMID
PubMedmcp__claude_ai_PubMed__find_related_articlesRelated articles for a PMID
PubMedmcp__claude_ai_PubMed__lookup_article_by_citationVerify a citation
PubMedmcp__claude_ai_PubMed__convert_article_idsConvert between PMID/DOI/PMCID
Semantic Scholarmcp__claude_ai_Scholar_Gateway__semanticSearchSemantic search across all fields
bioRxiv/medRxivmcp__claude_ai_bioRxiv__search_preprintsSearch preprint servers
bioRxiv/medRxivmcp__claude_ai_bioRxiv__get_preprintFull preprint metadata
CrossRefWebFetch with https://api.crossref.org/works/{DOI}DOI verification
\n

Fallback: NCBI E-utilities (Direct API via Bash)

\n

When PubMed MCP is unavailable (session timeout, \"MCP session has been terminated\" error,\nor \"No such tool available\" error), fall back to NCBI E-utilities via bundled scripts.

\n

Detection: If any mcp__claude_ai_PubMed__* call returns an error containing\n\"terminated\", \"not found\", \"not available\", or \"not connected\", switch ALL subsequent\nPubMed calls in this session to E-utilities. Do not retry MCP after a disconnect — it\nwill not recover within the same conversation.

\n

Scripts (in ${CLAUDE_SKILL_DIR}/references/):

\n\n

Usage patterns:

\n
EUTILS=\"${CLAUDE_SKILL_DIR}/references/pubmed_eutils.sh\"\nPARSER=\"${CLAUDE_SKILL_DIR}/references/parse_pubmed.py\"\n\n# Search PubMed (returns PMIDs)\nbash \"$EUTILS\" search \"diagnostic test accuracy meta-analysis radiology\" 20 \\\n  | python3 \"$PARSER\" esearch\n\n# Get article summaries as markdown table\nbash \"$EUTILS\" fetch_json \"16168343,16085191,31462531\" \\\n  | python3 \"$PARSER\" esummary\n\n# Get detailed metadata\nbash \"$EUTILS\" fetch \"16168343\" \\\n  | python3 \"$PARSER\" efetch\n\n# Generate BibTeX entries\nbash \"$EUTILS\" fetch \"16168343,16085191\" \\\n  | python3 \"$PARSER\" bibtex\n\n# Verify a citation by exact title\nbash \"$EUTILS\" cite_lookup \"Bivariate analysis of sensitivity and specificity\" \\\n  | python3 \"$PARSER\" esearch\n\n# Find related articles for a PMID\nbash \"$EUTILS\" related \"16168343\" 10 \\\n  | python3 \"$PARSER\" esummary\n
\n

Rate limiting: 3 requests/second without API key, 10/sec with NCBI_API_KEY.\nThe script auto-sleeps 350ms between calls. For batch operations, keep calls sequential.

\n

E-utilities → MCP equivalence:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
MCP ToolE-utilities CommandParser Mode
search_articlessearch <query> [retmax]esearch
get_article_metadatafetch <pmids>efetch or bibtex
find_related_articlesrelated <pmid> [retmax]esummary
lookup_article_by_citationcite_lookup <title>esearch → fetch
convert_article_idsNot available (use CrossRef DOI lookup)—
\n
\n

Workflow

\n

Phase 1: Search Strategy

\n
    \n
  1. Understand the need: Get the research topic, specific question, or manuscript section\nthat needs references.
    2. \n
  2. Generate search terms:\n\n
    3. \n
  3. Define scope:\n\n
    4. \n
  4. Present the search plan to the user before executing. Include the Boolean query,\ndatabases to search, and filters.
    5. \n
\n

Gate: Wait for user approval before running searches.

\n

Phase 2: Execute Search

\n
    \n
  1. Search PubMed using search_articles with the Boolean query.
    2. \n
  2. Search Semantic Scholar using semanticSearch with natural language query.
    3. \n
  3. Search bioRxiv/medRxiv using search_preprints if preprints are relevant.
    4. \n
  4. Deduplicate results across databases (match by DOI or title similarity).
    5. \n
  5. Present results in a structured table:
    6. \n
\n
| # | Title | Authors (first + last) | Year | Journal | PMID/DOI | Relevance |\n|---|-------|----------------------|------|---------|----------|-----------|\n| 1 | ...   | Kim J, ... Lee S     | 2024 | Radiology | 12345678 | High      |\n
\n
    \n
  1. Ask the user to select which papers to include.
    2. \n
\n

Record what the source said existed, not only what you downloaded

\n

Search code reports its own haul. Nothing errors when the haul is wrong, and a PRISMA flow built on\na wrong number is fiction that nothing downstream contradicts. Two signatures, both real, both from\na single run:

\n\n

PubMed via E-utilities is the one place where the naive pattern happens to be safe. Everywhere\nelse, do both.

\n

A DOI in a screening row is not necessarily that row's DOI

\n

When a doi column was filled by the pipeline rather than handed over with the record — matched\nagainst Crossref by title similarity, at some threshold — a wrong match is a valid, resolvable\nidentifier for a different paper, and nothing downstream can tell. Resolve it and read the title\nback before any decision rests on it:

\n
python3 scripts/check_doi_record_match.py --table 2_Screening/round3.tsv \\\n        --email <contact> --json qc/doi_record_match.json\n
\n

DOI_NOT_THIS_RECORD is a DOI that resolves to another paper; DOI_IS_CONTAINER is one that\nresolves to an issue, supplement or proceedings rather than an article; DOI_UNRESOLVED is\nreported rather than dropped. This is not /verify-refs, which audits a finished reference list —\nit runs at screening, where a wrong DOI is still cheap. In one review two of these appeared within\ntwo days, and one produced a limitation about a \"missed eligible paper\" that did not exist.

\n

Phase 2.5: Citation Searching (Snowballing)

\n

Optional but recommended for systematic reviews and thorough background work\n(PRISMA item 7, \"records identified through citation searching\"). Expands a\nseed set along the citation graph instead of relying on Boolean recall alone.

\n

Use the deterministic helper references/snowball.py (Semantic Scholar Graph\nAPI; nothing generated from memory):

\n
# Expand seed DOIs/PMIDs in all directions, dedup against the existing pool,\n# append verified candidates to references/library.bib\npython3 references/snowball.py \\\n  --seed DOI:10.1148/radiol.2024123,PMID:38000001 \\\n  --direction all \\\n  --pool references/library.bib \\\n  --out references/library.bib\n
\n\n

A deterministic, network-free challenge card (recorded fixtures + expected\noutput + verify.sh) lives in references/snowball_challenge/.

\n

Phase 3: Deep Read

\n

For each selected paper:

\n
    \n
  1. Retrieve full metadata using get_article_metadata (PubMed) or get_preprint (bioRxiv).
    2. \n
  2. Extract key information:\n\n
    3. \n
  3. Build a literature matrix if multiple papers selected:
    4. \n
\n
| Paper | Design | N | Key Finding | Limitation | Relevance to Our Study |\n|-------|--------|---|-------------|------------|----------------------|\n
\n
    \n
  1. Present the matrix to the user for review.
    2. \n
\n

Phase 4: Citation Management

\n

Anti-Hallucination Protocol

\n

This is the most critical part of the skill. Follow these rules without exception:

\n
    \n
  1. NEVER generate a reference from memory alone. Every reference must come from an API search result.
    2. \n
  2. NEVER fabricate DOIs or PMIDs. If you cannot find a DOI/PMID, mark the reference as [UNVERIFIED - NEEDS MANUAL CHECK].
    3. \n
  3. Cross-check every reference against the API result:\n\n
    4. \n
  4. If any field does not match, flag the specific mismatch.
    5. \n
  5. For DOI verification, use WebFetch with https://api.crossref.org/works/{DOI} to confirm the DOI resolves correctly.
    6. \n
\n

BibTeX Generation

\n

For each reference (verified or not), generate a BibTeX entry with an explicit\nverified flag so downstream skills (/lit-sync, /verify-refs,\n/write-paper) can reason about trust without re-running verification:

\n
@article{FirstAuthorLastName_Year_ShortKey,\n  author    = {Last1, First1 and Last2, First2 and Last3, First3},\n  title     = {Full Title As Retrieved From Database},\n  journal   = {Journal Name},\n  year      = {2024},\n  volume    = {310},\n  number    = {2},\n  pages     = {e234567},\n  doi       = {10.1001/jama.2024.12345},\n  pmid      = {12345678},\n  verified  = {true},\n  verified_by = {pubmed+crossref},\n  verified_on = {2026-04-24},\n}\n
\n

verified flag values (required on every entry):

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ValueMeaningDownstream behavior
trueDOI or PMID confirmed via PubMed/CrossRef; title, authors, year all matchSafe to cite; /write-paper citekey-only gate passes
falseParsed from text but API lookup failed or returned mismatch/verify-refs flags as UNVERIFIED; manuscript MUST show [UNVERIFIED - NEEDS MANUAL CHECK]
manualUser explicitly added despite lookup failureTreated as verified=false by /verify-refs but suppresses repeat warnings
\n

verified_by lists the data sources that confirmed the entry (e.g., pubmed,\ncrossref, semantic_scholar, or a combination). verified_on is the ISO date\nof the most recent successful verification.

\n

BibTeX key convention: FirstAuthorLastName_Year_OneWord (e.g., Kim_2024_Validation).

\n

Output

\n
    \n
  1. Save BibTeX entries to the specified .bib file (append, do not overwrite).\nTarget: references/library.bib (candidate pool for /lit-sync to import\ninto Zotero). NEVER write to manuscript/_src/refs.bib — that is /lit-sync's\nsole-writer path per docs/artifact_contract.md.
    2. \n
  2. Print a summary of all references with verification status:
    3. \n
\n
Verified:    12 references (verified=true)\nUnverified:   1 reference  (verified=false) [NEEDS MANUAL CHECK]\nTotal:       13 references\n
\n

Phase 4b: Zotero Library Integration

\n

If a Zotero MCP server is available, integrate search results with the user's library:

\n
    \n
  1. Check for duplicates first: Use zotero_search_items (by DOI) to skip papers already in the library — this search-first step is what dedupes; zotero_add_by_doi does not dedupe on its own.
    2. \n
  2. Add papers to Zotero: Use zotero_add_by_doi for DOI-based import (its attach_mode argument governs the OA PDF attach attempt at add time).
    3. \n
  3. Organize into collections: Use zotero_manage_collections to file into the relevant project collection.
    4. \n
  4. Leverage annotations: Use zotero_get_annotations to reference the user's prior reading notes.
    5. \n
  5. Write sync audit: Record collection key, added/skipped/failed counts, and\nunsynced entries in references/zotero_collection.json so Zotero status is\nauditable rather than a hidden optional side effect.
    6. \n
\n
\n

Requires Zotero Desktop running with MCP server. Skip this phase if unavailable.\nIf skipped, still write references/zotero_collection.json with\nstatus: \"skipped\" and the reason.

\n
\n

Phase 5: Full-Text Retrieval

\n

Full-text PDF retrieval is delegated to /fulltext-retrieval — the single authored\nhome of the open-access cascade (arXiv → Unpaywall → PMC → OpenAlex → Crossref → landing\npage, each validated with a %PDF- header + ≥10 KB size). Do not re-implement OA\nfetching here.

\n

Pass the verified candidate DOIs from references/library.bib:

\n
ENGINE=\"${MEDSCI_SKILLS_ROOT:-$HOME/workspace/medsci-skills}/skills/fulltext-retrieval/fetch_oa.py\"\n# extract DOI + Title (and PMID/FirstAuthor when available) → worklist.tsv\npython3 \"$ENGINE\" worklist.tsv -o pdfs/ -e <contact-email> --report pdfs/retrieval_report.json\n
\n

Use the verified bibliographic title rather than discarding it into a DOI-only list.\nKeep source_identity and file_sha256 from the retrieval report with the record.\nDownload success and title agreement alone do not verify the PDF: inspect conflicts,\nunresolved/unavailable evidence, and files whose hashes have changed before citing them.\nMissing identity fields in older reports mean unassessed. Even consistent is advisory\nfront-matter corroboration, not verification of the claims in the paper.

\n

For Zotero-resident PDFs and higher-yield, proxy-aware retrieval, use /lit-sync Phase 2.7,\nwhich also invokes /fulltext-retrieval and triggers Zotero's native \"Find Available PDF\".

\n

Alternative sources (legitimate only)

\n

For DOIs that open access cannot reach (listed in pdfs/manual_needed.txt):

\n\n

Never bypass paywalls or publisher access controls, and do not configure unauthorized\nPDF mirrors. Rate limits and PDF validation are handled inside /fulltext-retrieval.

\n

Phase 6: Gap Analysis

\n

When called during manuscript writing (especially by /write-paper Phase 7):

\n
    \n
  1. Read the manuscript to extract all inline citations.
    2. \n
  2. Compare cited references against the search results.
    3. \n
  3. Identify gaps:\n\n
    4. \n
  4. Report findings to the user with specific suggestions.
    5. \n
\n
\n

Specialized Search Modes

\n

Mode: Manuscript Paper Reference Pool

\n

For supplying a manuscript's reference pool — typically invoked by /write-paper Step 7.3c (or\n/self-review Phase 2.5c-2) when the reference adequacy gate finds the draft under target or a\nnamed method uncited, but usable directly when building out an original-research bibliography.

\n

This mode is deliberately broad: for an original-research article, return 25–40 verified\ncandidates, not the ~10 a quick search settles on. Do not stop early unless the field is genuinely\nsparse — and if it is, say so explicitly rather than returning a thin list silently. Respect a\nnarrower journal reference cap or user scope when one is given.

\n

Structure the pool across six candidate categories so the gaps the adequacy gate cares about\nare all covered:

\n
    \n
  1. Background / disease burden / clinical context — establishes why the question matters.
    2. \n
  2. Gap-defining prior studies — the work the manuscript extends or contradicts.
    3. \n
  3. Comparator / comparable-design cohorts — studies the Results will be measured against.
    4. \n
  4. Methods / statistical canonical sources — the originating reference for every named method,\nmodel, score, equation, or diagnostic criterion (e.g. competing-risk model, multiple\nimputation, E-value, eGFR equation, concordance statistic). This is the category that clears\nMethods named-method gaps.
    5. \n
  5. Reporting-guideline sources — STROBE, TRIPOD(+AI), CONSORT, PRISMA(-DTA), STARD, etc.
    6. \n
  6. Interpretation / mechanism / limitation support — grounds Discussion claims.
    7. \n
\n

For each candidate, report: PMID/DOI, verification status, candidate category, the\ntarget manuscript section it belongs in, and a one-line why it is needed.

\n

Boundary (unchanged): every entry is API-verified before inclusion, and BibTeX is appended only\nto references/library.bib — the candidate pool for /lit-sync to import into Zotero. Never\nwrite to manuscript/_src/refs.bib; that SSOT belongs to /lit-sync. This mode produces\ncandidates; it does not decide inclusion (the user does) and it does not insert references into the\nmanuscript bib.

\n

Mode: Crowding Check

\n

Run before a study is designed, not after. The question is not \"what has been written about this\ntopic\" — a background search answers that and still leaves the trap open. It is narrower and it is\nfour questions:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Ask ofVerdict
the research questiontaken / partly taken / open
the sampling frame (what population, which records, which years)taken / partly taken / open
the measurement axis (what is being coded or measured, and at what granularity)taken / partly taken / open
the target journalalready published there / adjacent / open
\n

Each gets its own verdict. A design can be original on one axis and fully occupied on another, and\ncollapsing the four into one answer is what hides that.

\n

Why the fourth row is not vanity: a design once matched an existing paper on frame, coding axis\nand target journal, and that paper was already published in the journal it was first choice\nfor. A redesign on a different axis then turned out to be partly occupied too — three papers were\nalready coding the same thing as a single item — which did not kill it but did change the claim\nthat could honestly be made, from \"nobody has looked at this\" to \"nobody has decomposed it by\nprovenance\". That is a real result of this mode: most of the time it narrows a claim rather than\nending a project, and a narrowed claim survives review where the broad one would not.

\n

Search the way a competitor would: the exact frame, the exact measure, and the journal's own site,\nnot only the topic. Report the four verdicts and the papers behind each, then let the user decide.\n/design-study and /orchestrate should route here first when a new study is being scoped.

\n

Mode: Systematic Search

\n

For systematic reviews or comprehensive literature sections:

\n
    \n
  1. Document the full search strategy (PRISMA-compliant).
    2. \n
  2. Record: database, date of search, query string, number of results.
    3. \n
  3. Track inclusion/exclusion at each screening step.
    4. \n
  4. Output a PRISMA flow diagram data summary.
    5. \n
\n

Mode: Quick Cite

\n

For quickly finding a single reference the user describes:

\n
    \n
  1. User says something like \"that 2023 paper by Smith about AI in chest X-ray.\"
    2. \n
  2. Search PubMed and Semantic Scholar with the described details.
    3. \n
  3. Present top 3 candidates.
    4. \n
  4. User confirms which one.
    5. \n
  5. Generate BibTeX entry.
    6. \n
\n

Mode: Related Papers

\n

For expanding from a known paper:

\n
    \n
  1. User provides a PMID or DOI.
    2. \n
  2. Use find_related_articles to get related papers.
    3. \n
  3. Use Semantic Scholar for citation-based recommendations.
    4. \n
  4. Present results ranked by relevance.
    5. \n
\n

For a structured, dedup-aware, PRISMA-countable expansion (backward +\nforward + similar) prefer Phase 2.5: Citation Searching with\nreferences/snowball.py, which appends verified candidates to\nreferences/library.bib and reports a citation-searching count.

\n

Mode: Embase Browser Automation

\n

Embase has no public API. Use Chrome browser automation (MCP) to search and export:

\n
    \n
  1. Navigate to embase.com — institutional SSO authenticates automatically.\nIf cookie error (login?error#), clear Elsevier/Embase cookies and retry.
    2. \n
  2. Go to Advanced Search tab.
    3. \n
  3. Enter Embase-syntax query (Emtree /exp + :ab,ti field tags).\nUncheck \"Map to preferred term in Emtree\" when using explicit /exp terms.
    4. \n
  4. After results appear, use \"Select number of items\" dropdown → select total count.
    5. \n
  5. Click Export (in Results section) → choose CSV format → check fields:\nTitle, Author names, Source, Publication year, Publication type, DOI, Abstract,\nLanguage of article, Medline PMID.
    6. \n
  6. Click Export → Download tab opens → click Download.
    7. \n
  7. CSV is in row format (records separated by blank rows) — parse with:\n
    # Each record = consecutive rows until blank row\n# Row format: [FIELD_NAME, value1, value2, ...]\n# AUTHOR NAMES row has multiple values (one per author)\n
    \n
    8. \n
\n

PubMed → Embase query translation:

\n\n
\n

Error Handling

\n\n

What This Skill Does NOT Do

\n\n","files":[{"path":"references/embase_browser.md","sizeBytes":1351,"isText":true},{"path":"references/parse_pubmed.py","sizeBytes":14111,"isText":true},{"path":"references/pubmed_eutils.sh","sizeBytes":4295,"isText":true},{"path":"references/snowball_challenge/expected/snowball.bib","sizeBytes":1576,"isText":false},{"path":"references/snowball_challenge/fixture/DOI_10_0_seed1.backward.json","sizeBytes":600,"isText":true},{"path":"references/snowball_challenge/fixture/DOI_10_0_seed1.forward.json","sizeBytes":571,"isText":true},{"path":"references/snowball_challenge/fixture/DOI_10_0_seed1.similar.json","sizeBytes":263,"isText":true},{"path":"references/snowball_challenge/fixture/DOI_10_0_seed_err.backward.json","sizeBytes":59,"isText":true},{"path":"references/snowball_challenge/fixture/DOI_10_0_seed_trunc.forward.json","sizeBytes":678,"isText":true},{"path":"references/snowball_challenge/fixture/library.bib","sizeBytes":277,"isText":false},{"path":"references/snowball_challenge/problem.md","sizeBytes":2687,"isText":true},{"path":"references/snowball_challenge/verify.sh","sizeBytes":3125,"isText":true},{"path":"references/snowball.py","sizeBytes":14990,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Farticle.008.json","sizeBytes":278,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Ferratum.007.json","sizeBytes":292,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Fmarkup.005.json","sizeBytes":180,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Fmatch.001.json","sizeBytes":183,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Fother.002.json","sizeBytes":154,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Fsuppl.003.json","sizeBytes":135,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/cache/10.1000%2Fversion.009.json","sizeBytes":265,"isText":true},{"path":"scripts/check_doi_record_match_challenge/fixture/screening.tsv","sizeBytes":615,"isText":false},{"path":"scripts/check_doi_record_match_challenge/fixture/update_notice.tsv","sizeBytes":444,"isText":false},{"path":"scripts/check_doi_record_match_challenge/verify.sh","sizeBytes":6065,"isText":true},{"path":"scripts/check_doi_record_match.py","sizeBytes":13230,"isText":true},{"path":"SKILL.md","sizeBytes":18933,"isText":true},{"path":"skill.yml","sizeBytes":2909,"isText":true},{"path":"tests/test_parse_pubmed.sh","sizeBytes":3967,"isText":true},{"path":"tests/test_pubmed_eutils.sh","sizeBytes":3775,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"notes-only","suspicious":0,"notes":1,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-10-10T01:48:30.619646Z","sha256":"EC028A597B404B4102405C0BE55C6579BB36A49CC2E841C4E7167755A5A98504","sizeBytes":43325},"review":null,"source":{"repositoryUrl":"https://github.com/Aperivue/medsci-skills","path":"skills/search-lit","license":"MIT","commit":"3b14ae2a9424a3a2b14f38338bc527823df06e96","subtreeSha":"41ED96E9281D396ED3B0B73FC122367498D8100C67FE67C19950286C9B64E7EC","lastSyncedAt":"2026-10-10T01:47:39.165903Z"},"reviewedAt":"2026-10-10T01:50:17.002951Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/Aperivue/medsci-skills/tree/main/skills/search-lit"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install aperivue-medsci-skills@llmmart"},{"target":"git","command":"git clone https://github.com/Aperivue/medsci-skills.git"}]}