More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Citation Management: Clean Reference Pipeline

    1

    Run a clean reference pipeline for research writing: discover across databases, pull canonical metadata, keep one trustworthy BibTeX source, and validate on a schedule — no archaeology at submission.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Citation Management: Clean Reference Pipeline

    Citation Management: Clean Reference Pipeline

    Example session with this skill installed

    Three co-authors, three divergent bibliography files, 60 entries total, submission in six weeks. Clean it up and make it stay clean.

    • Read your context and instructions
    • Compiled the citation management

    ONE FILE: The three files merge into one canonical bibliography under version control. The merge alone surfaces 9 duplicates by DOI — 15% of the file.

    RE-FETCH: Every entry re-fetched from its canonical source and diffed. Twelve entries have wrong years (arXiv year used for published papers), four have swapped author order, three are missing venues, two cite workshop papers as conference papers. All fixed from the canonical records.

    KEYS: Keys normalize to author-year-keyword. Eleven citations in the manuscript point at old keys; a script finds and fixes each one. No orphaned keys remain.

    VALIDATION: Checklist runs: every identifier resolves, every entry is cited except two background entries (marked and excluded from the submitted bibliography), formatting consistent after one pass on venue abbreviations.

    RESULT: The bibliography ships with 51 entries, down from 60, every one resolving, complete, and cited. The validation routine is scheduled to run on every bibliography change, so the file never rots again. Total cleanup: two focused sessions, and the pipeline that prevents the next one. The cleanup report records the nine duplicates removed, the eighteen metadata fixes with their canonical sources, and the eleven key corrections, so the next author can see exactly what changed and why.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Reference rot is gradual: a duplicated entry here, a wrong year there, three divergent bibliography files across co-authors. By submission time, cleaning it up is archaeology. This skill sets up the pipeline so the bibliography stays clean from the first paper you read: search more than one database (OpenAlex for breadth, PubMed for biomedical, CrossRef for DOI metadata, arXiv for preprints — discover anywhere, verify against a canonical source before entry), capture metadata at first contact with a permanent identifier and a one-line "why it matters," and keep one canonical BibTeX file under version control with stable author-year-keyword keys.

    Validation runs on a schedule, not just at submission: every identifier resolves, no duplicates by DOI, every entry cited or marked background, every citation key matched, formatting consistent. An inherited mess gets cleaned once, in the right order — dedupe by DOI first (often 10–20%), re-fetch metadata, normalize keys, delete the uncited — and the pipeline prevents the next cleanup. A worked example takes a 60-entry three-file mess to 51 clean entries in two sessions.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    • Verification needs a canonical source; an entry that cannot be verified does not enter the library yet.
    • Preprint and published versions must be tracked separately — the skill cites the version actually read.
    • It manages references; it does not audit whether citations support their claims (that is the citation audit's job).
    • Co-authors editing outside version control will re-create divergence; the single file must stay the only source of truth.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    The manager is the storage; this is the discipline. One canonical file, stable keys, verified metadata at entry, validation on a schedule — without those, the manager just stores the rot more efficiently.

    Duplicates with slightly different keys. Two entries for the same paper, keyed differently, breed silently and surface at submission. The duplicate check by DOI before adding catches them.

    The pipeline failed, not the cleanup. Cleaning the same bibliography twice means there was no single file under version control, no validation schedule, or no verify-before-entry rule. Fix the pipeline and the cleanup stays done.

    New in LLM Ops