agentleFS
Durden-the-coderDurden-the-coder/Durden_skills/skills/ai-biomedical-journal-watch/SKILL.md

ai-biomedical-journal-watch

Skill4 starsChanged 3 days agoMITView on GitHub

Collect, screen, rank, audit, and summarize recent AI-related biomedical research and journal trend content from 18 configured journals. Supports customizable rolling windows, evidence-based decisions, prestige-aware ranking, bounded Chinese reports, and pre-render validation.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read references/version.json and use the declared 18-journal scope.

    line 8
  2. Very specific:

    Read references/time-windows.md, references/source-strategies.md, references/journal-priority.json, and references/screening-contract.md.

    line 12
  3. Distinctive:

    Run the mandatory audit before rendering:

    line 17
  4. Very specific:

    Do not render a final report when the audit exits nonzero.

    line 23
  5. Very specific:

    Treat a Letter as research only when abstract/lead evidence supports original methods or results.

    line 33
  6. Very specific:

    Do not infer AI from lexical collisions.

    line 34

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Biomedical Journal Watch v0.2.1

Read references/version.json and use the declared 18-journal scope. Clinical coverage is limited to NEJM and The Lancet.

Workflow

  1. Read references/time-windows.md, references/source-strategies.md, references/journal-priority.json, and references/screening-contract.md.
  2. Freeze as_of; discover from preferred sources and fallbacks; merge by DOI, PMID, then normalized title.
  3. Store every candidate before screening. For windows longer than seven days, collect in seven-day slices and screen in batches of 40.
  4. Preserve publication date, article type, abstract/lead evidence, and evidence URLs.
  5. Assign include, context, exclude, or needs_evidence per record. Never implement decisions as DOI/title sets, dictionaries, tuples, whitelists, blacklists, index tables, or summary override tables in executable code.
  6. Run the mandatory audit before rendering:

    python <skill-root>\scripts\audit_reviewed_jsonl.py --input <reviewed.jsonl> --output <audit.json>

    Do not render a final report when the audit exits nonzero. Review warnings in large runs, especially zero needs_evidence.

  7. Render with the window-aware stable entry point:

    python <skill-root>\scripts\render_report_v321.py --window 30d --input <reviewed.jsonl> --output <report.md>

Screening guardrails

  • Route Review, Perspective, Preview, Editorial, News, Highlight, Comment, Commentary, and Viewpoint to context when AI-relevant; never count them as original research.
  • Treat a Letter as research only when abstract/lead evidence supports original methods or results. Otherwise use needs_evidence or context.
  • Do not infer AI from lexical collisions. Names such as transformer base editor, “intelligent” materials, predictive biomarkers, automated assays, Bayesian/statistical models, or ordinary computational pipelines are not AI evidence by themselves.
  • Require an explicit evidenced AI/ML function: training, learned inference, model evaluation, representation/generative learning, or an evaluated AI intervention.
  • Prestige affects ranking only, never eligibility.
  • Included/context items require dates and evidence URLs. Included items require supported summaries.

Output policy

  • Detailed research: 20 for every window; per-journal tier cap never above 5.
  • Additional research index: 30 for 72h/96h, 50 for 7d, 80 for 30d or longer.
  • Context list: 20 by default; title, journal, date, type, and link only.
  • Exclusion summary: top 10 normalized categories by default; keep full reasons in JSONL.
  • Markdown is written with a UTF-8 BOM for Windows compatibility. JSONL remains UTF-8.

Source exceptions

  • Nature Communications: enumerate and paginate the official article list; RSS alone is insufficient.
  • Nature Methods: follow official pages for abstract/date/type.
  • Cell Metabolism: inspect PubMed/Europe PMC publication types and correction links before enrichment.
  • Cell Press, AAAS, and The Lancet: when publisher pages block access, use the Crossref/PubMed/Europe PMC union.
  • NEJM: use RSS and PubMed, then OpenAlex for missing abstracts.

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.