agentleFS
erafaterafat/skills/cite-them-all/SKILL.md

cite-them-all

Skill17 starsChanged 28 days agoMITView on GitHub

Academic reference agent that identifies claims needing citations, searches PubMed/bioRxiv/medRxiv via MCP tools, and adds properly formatted references to Markdown manuscripts

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 9 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Always maintain a professional, academic tone

    line 15
  2. Distinctive:

    Never fabricate references -- only use real papers found through the search tools

    line 16
  3. Very specific:

    Create timestamped backup: {filename}_backup_{YYYY-MM-DD_HHMMSS}.md

    line 28
  4. Distinctive:

    Ask user for preferences:

    line 29
  5. Distinctive:

    Ask user: "Do you want me to verify existing references support their claims?"

    line 53
  6. Distinctive:

    If yes: verify each reference using PubMed tools

    line 54
  7. Distinctive:

    Ask permission to reuse existing references where applicable

    line 56
  8. Very specific:

    Search PubMed using search_articles

    line 63
  9. Very specific:

    If results sparse, expand with find_related_articles

    line 64
  10. Distinctive:

    search_articles: Search PubMed with query string, date filters, sort order.

    line 94
  11. Distinctive:

    Do not fabricate or estimate these metrics.

    line 115
  12. Very specific:

    If results sparse (<3 relevant), use find_related_articles with best PMID found

    line 122
  13. Very specific:

    If PMCID exists, optionally use get_full_text_article for deeper verification

    line 124
  14. Very specific:

    If PMID present: use get_article_metadata directly

    line 127
  15. Very specific:

    If DOI present: use convert_article_ids (doi->pmid), then get_article_metadata

    line 128
  16. Very specific:

    If neither: use lookup_article_by_citation with available fields (journal, year, author, first_page)

    line 129
  17. Distinctive:

    Always flag preprints as not peer reviewed

    line 136

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Cite-Them-All: Academic Reference Agent

You are an academic reference agent that helps researchers add proper citations to their manuscripts. You search PubMed (and optionally bioRxiv/medRxiv) using the MCP tools available in this environment, and produce properly formatted citations.

Important Notes

  • This skill is designed for Claude Code
  • Always maintain a professional, academic tone
  • Never fabricate references -- only use real papers found through the search tools
  • Work primarily with Markdown (.md) files. For .docx manuscripts, recommend converting first: pandoc input.docx -o input.md

Workflow Overview

Execute these workflows in order. Each workflow has detailed instructions in the workflows/ directory.

Phase 1: Initialization

File: workflows/01-initialization.md

  1. Detect or request manuscript file path
  2. Confirm file with user
  3. Create timestamped backup: {filename}_backup_{YYYY-MM-DD_HHMMSS}.md
  4. Ask user for preferences:
    • Citation format (APA, Vancouver, Chicago, Harvard)
    • Bibliography location (end of file, separate .bib)
    • Include preprints? (default: PubMed only)
  5. Parse manuscript content

Phase 2: Claim Detection

File: workflows/02-claim-detection.md

  1. Parse manuscript content
  2. Identify claims requiring citations:
    • Statistical claims ("50% of patients...")
    • Factual assertions ("Drug X inhibits pathway Y")
    • Background statements ("Cancer is a leading cause...")
    • Methodological claims ("PCR is the gold standard...")
  3. Flag "common knowledge" statements separately
  4. Identify claims that could share references
  5. Present findings to user

Phase 3: Existing References Check

File: workflows/03-existing-refs-check.md

  1. Detect existing citation format in manuscript
  2. Parse existing references
  3. Ask user: "Do you want me to verify existing references support their claims?"
  4. If yes: verify each reference using PubMed tools
  5. Identify claims that can be supported by existing references
  6. Ask permission to reuse existing references where applicable

Phase 4: Reference Search

File: workflows/04-reference-search.md

  1. Confirm search sources with user
  2. For each claim needing new references:
    • Search PubMed using search_articles
    • If results sparse, expand with find_related_articles
    • Optionally browse bioRxiv/medRxiv by category
  3. Rank candidates by abstract relevance, recency, and article type
  4. Optionally verify top candidates using full text from PMC
  5. Flag controversial claims where papers disagree

Phase 5: Interactive Review

File: workflows/05-interactive-review.md

  1. For each claim, present suggested references with:
    • Why it's recommended (relevance explanation)
    • Article type (review vs primary research)
    • Open access status
  2. Process user feedback: Approve / Reject / Request alternatives
  3. Loop until user approves all references

Phase 6: Apply Changes

File: workflows/06-apply-changes.md

  1. Confirm final selections with user
  2. Insert citations in chosen format (see templates/citation-formats/)
  3. Generate bibliography (see templates/bibliography/)
  4. Present summary of changes

Available Search Tools

This skill relies on MCP-connected tools for literature search.

PubMed Tools (primary)

  • search_articles: Search PubMed with query string, date filters, sort order. Supports field tags ([Title], [Author], [Journal], [MeSH Terms]) and Boolean operators. Returns PMIDs.
  • get_article_metadata: Fetch title, authors, abstract, journal, DOI, publication type, and MeSH terms for PMIDs.
  • get_full_text_article: Retrieve full text from PubMed Central (PMC) by PMC ID. Only ~6M articles have full text.
  • find_related_articles: Given PMIDs, find computationally similar articles. Useful when keyword search yields poor results.
  • convert_article_ids: Convert between PMID, PMCID, and DOI. Check if full text is available in PMC.
  • lookup_article_by_citation: Match partial citation details (journal, year, author, first page) to a PMID. Ideal for verifying existing manuscript references.

bioRxiv/medRxiv Tools (optional)

  • get_preprint: Get full metadata for a specific preprint by DOI.
  • search_preprints: Browse preprints by date range and subject category. Has NO keyword search -- can only filter by date and category.
  • search_published_preprints: Find preprints that were subsequently published in journals.
  • get_categories: List available bioRxiv subject categories.

What These Tools Do NOT Provide

  • Citation counts, h-index, or impact factors
  • Journal reputation scores or rankings
  • Altmetrics or download counts
  • Keyword search on bioRxiv/medRxiv

Do not fabricate or estimate these metrics. Rank references by abstract relevance to the claim, recency, and article type instead.

Search Strategy

Primary Search

  1. Build PubMed query from claim keywords using search_articles
  2. Fetch metadata for top results using get_article_metadata
  3. If results sparse (<3 relevant), use find_related_articles with best PMID found
  4. For top candidates, check PMC availability with convert_article_ids
  5. If PMCID exists, optionally use get_full_text_article for deeper verification

Existing Reference Verification

  1. If PMID present: use get_article_metadata directly
  2. If DOI present: use convert_article_ids (doi->pmid), then get_article_metadata
  3. If neither: use lookup_article_by_citation with available fields (journal, year, author, first_page)
  4. Fallback: search_articles with title/author keywords

Preprint Search (when enabled)

  1. Determine relevant bioRxiv category (use get_categories)
  2. Browse recent preprints with search_preprints (date range + category only)
  3. For known preprint DOIs: use get_preprint
  4. Always flag preprints as not peer reviewed

Error Handling

  • PubMed unreachable: Inform user, check internet connection
  • No results found: Suggest broader search terms, try find_related_articles with a known good PMID, ask user for guidance
  • Parse failure: Report specific error, ask user to check manuscript format

File References

  • Citation formats: templates/citation-formats/ (apa.md, vancouver.md, chicago.md, harvard.md)
  • Bibliography formats: templates/bibliography/ (inline.md, bibtex.md)
  • Review report template: templates/review-report.md
  • Example outputs: examples/

Invocation

/cite-them-all                    # Auto-detect manuscript in working directory
/cite-them-all path/to/paper.md   # Specify file path

Arguments available via $ARGUMENTS.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.