ncbi-sequence-fetch

name	ncbi-sequence-fetch
description	Retrieve protein and nucleotide sequences from NCBI via E-utilities: by accession, CDS translation, gene+organism search, locus lookup, PubMed-linked sequences, or patent proteins. Use to fetch biological sequences. Public NCBI API, no credential needed.
license	Apache-2.0

Ncbi Sequence Fetch (Codex-native)

Codex-native adaptation of Google DeepMind's ncbi-sequence-fetch skill (science-skills, Apache-2.0). Uses the public API directly through Codex's own tools.

Gates (ask before proceeding)

• Terms notice: on first use in a workspace, tell the user to review the source's terms of use and record the notice in provenance.
• Network / install: read-only public access; be polite about rate limits. Write large responses to a file and parse them; if you write helper code, run it via uv run, never bare python3. Ask before installs or large downloads.

Endpoints

• NCBI E-utilities https://eutils.ncbi.nlm.nih.gov/entrez/eutils (esearch/efetch, db=protein|nuccore)

Workflow

1. For an accession, efetch with rettype=fasta; for a gene+organism, esearch then efetch.
2. Write large sequence output to a file; do not dump long FASTA to stdout.

• Provenance & review: record queries, endpoints, and cited identifiers with $science-provenance; check claims with $science-review before presenting.

Boundaries

• Report the accession and database (protein vs nucleotide); respect NCBI rate limits.
• Cite only accessions actually returned; never fabricate sequences.