From DOI to Downloaded Data: article-data-fetcher

Where this fits

A researcher reads a paper with a GWAS or single-cell result and wants to reproduce it. The VCF or H5AD file is not in the paper; it is hidden in GEO, ENA, Zenodo, Figshare, Dryad, or OSF. Instead of manually clicking through accession numbers, an agent can use article-data-fetcher to take the article DOI/PMID and download the deposited files directly.

Why agents benefit

• Resolves article identifiers to repository accessions without hand-curated mappings.
• Queries multiple public repositories in one pass, reducing broken-link dead ends.
• Filters by file types (vcf, fasta, h5ad, csv, bam) so agents pull only what the downstream task needs.
• Saves outputs into a structured local directory, making pipeline handoffs predictable.
• Works as a discrete skill, so agent workflows stay modular and testable.

Practical scenarios

Reproducing a GWAS from a published VCF

A bioinformatics agent is asked to validate a paper's findings. It takes the DOI, requests a vcf file, and article-data-fetcher locates the summary statistics and genotype calls in ENA/GEO, downloads them, and hands the VCF to a QC step.

Preparing single-cell data for re-analysis

A user wants to reanalyze a single-cell experiment but only has the paper's PMID. The agent asks for h5ad files, the skill retrieves the AnnData object from Zenodo/Figshare, and the agent can immediately start clustering or differential expression.

Building a variant-calling benchmark set

For a benchmarking task, an agent gathers multiple articles and downloads matching FASTA/FASTQ/BAM files. Because the skill accepts a list of extensions, the agent can assemble a ground-truth dataset without manually mapping each paper to a repository.

Add it to your agent workflow

Install the skill with:

npx skills add ClawBio/ClawBio --skill article-data-fetcher

Then instruct your agent to call it when a task names a paper or file type. A minimal example prompt:

Use article-data-fetcher with article_id=10.1038/s41588-024-01843-4,
file_types=vcf,fasta, output_dir=./downloads.

The skill resolves the DOI, finds files in public repositories, and returns the local paths as its output.

Compare before adopting

Before relying on it, compare against manual repository scripts or generic DOI-resolution APIs. Look at the coverage of repositories it actually reaches (GEO, ENA, Zenodo, Figshare, Dryad, OSF) and whether it handles non-open access gracefully. Check when the skill was last updated, how it handles failed downloads, and whether it fits into your existing agent loop. Alternatives like Entrez E-utilities or custom fetch scripts may be better if you need deeper control.

Why it is worth tracking

The skill solves a common but painful reproducibility gap: getting data from a paper into a local environment without human digging. With 1,125 stars and active maintenance from ClawBio, it has real community momentum and a focused scope. Evaluate it when you build agents that process papers, run meta-analyses, or answer data-availability questions. For a current assessment and skill card, see https://www.openagentskill.com/skills/clawbio-article-data-fetcher.