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
In this repository11 other files it gives its agents.
- anatomySkill
- anything-to-notebooklmSkill
- cv-update-reviewSkill
- illustration-productionSkill
- manuscript-productionSkill
- md-to-xhs-cardsSkill
- scientific-schematicsSkill
- scientific-slidesSkill
- scientific-visualizationSkill
- scientific-writingSkill
- xhs-note-creatorSkill
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
- Distinctive: line 15
Always maintain a professional, academic tone
- Distinctive: line 16
Never fabricate references -- only use real papers found through the search tools
- Very specific: line 28
Create timestamped backup: {filename}_backup_{YYYY-MM-DD_HHMMSS}.md
- Distinctive: line 29
Ask user for preferences:
- Distinctive: line 53
Ask user: "Do you want me to verify existing references support their claims?"
- Distinctive: line 54
If yes: verify each reference using PubMed tools
- Distinctive: line 56
Ask permission to reuse existing references where applicable
- Very specific: line 63
Search PubMed using search_articles
- Very specific: line 64
If results sparse, expand with find_related_articles
- Distinctive: line 94
search_articles: Search PubMed with query string, date filters, sort order.
- Distinctive: line 115
Do not fabricate or estimate these metrics.
- Very specific: line 122
If results sparse (<3 relevant), use find_related_articles with best PMID found
- Very specific: line 124
If PMCID exists, optionally use get_full_text_article for deeper verification
- Very specific: line 127
If PMID present: use get_article_metadata directly
- Very specific: line 128
If DOI present: use convert_article_ids (doi->pmid), then get_article_metadata
- Very specific: line 129
If neither: use lookup_article_by_citation with available fields (journal, year, author, first_page)
- Distinctive: line 136
Always flag preprints as not peer reviewed
≈ “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.
Files
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
- Detect or request manuscript file path
- Confirm file with user
- Create timestamped backup:
{filename}_backup_{YYYY-MM-DD_HHMMSS}.md - Ask user for preferences:
- Citation format (APA, Vancouver, Chicago, Harvard)
- Bibliography location (end of file, separate .bib)
- Include preprints? (default: PubMed only)
- Parse manuscript content
Phase 2: Claim Detection
File: workflows/02-claim-detection.md
- Parse manuscript content
- 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...")
- Flag "common knowledge" statements separately
- Identify claims that could share references
- Present findings to user
Phase 3: Existing References Check
File: workflows/03-existing-refs-check.md
- Detect existing citation format in manuscript
- Parse existing references
- Ask user: "Do you want me to verify existing references support their claims?"
- If yes: verify each reference using PubMed tools
- Identify claims that can be supported by existing references
- Ask permission to reuse existing references where applicable
Phase 4: Reference Search
File: workflows/04-reference-search.md
- Confirm search sources with user
- 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
- Search PubMed using
- Rank candidates by abstract relevance, recency, and article type
- Optionally verify top candidates using full text from PMC
- Flag controversial claims where papers disagree
Phase 5: Interactive Review
File: workflows/05-interactive-review.md
- For each claim, present suggested references with:
- Why it's recommended (relevance explanation)
- Article type (review vs primary research)
- Open access status
- Process user feedback: Approve / Reject / Request alternatives
- Loop until user approves all references
Phase 6: Apply Changes
File: workflows/06-apply-changes.md
- Confirm final selections with user
- Insert citations in chosen format (see
templates/citation-formats/) - Generate bibliography (see
templates/bibliography/) - 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
- Build PubMed query from claim keywords using
search_articles - Fetch metadata for top results using
get_article_metadata - If results sparse (<3 relevant), use
find_related_articleswith best PMID found - For top candidates, check PMC availability with
convert_article_ids - If PMCID exists, optionally use
get_full_text_articlefor deeper verification
Existing Reference Verification
- If PMID present: use
get_article_metadatadirectly - If DOI present: use
convert_article_ids(doi->pmid), thenget_article_metadata - If neither: use
lookup_article_by_citationwith available fields (journal, year, author, first_page) - Fallback:
search_articleswith title/author keywords
Preprint Search (when enabled)
- Determine relevant bioRxiv category (use
get_categories) - Browse recent preprints with
search_preprints(date range + category only) - For known preprint DOIs: use
get_preprint - 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_articleswith 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 pathArguments available via $ARGUMENTS.
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
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.

