Operating rules

• Use scripts/ncbi_entrez.py for all Entrez calls in this package.
• Use explicit endpoint values such as esearch, esummary, efetch, elink, or einfo.
• Search-style Entrez calls are better with retmax=10 and max_items=10.
• GEO is nested under this skill. Use db=gds or db=geoprofiles for GEO metadata and load references/geo.md only when the user is specifically asking about GEO.
• BLAST workflows belong in ncbi-blast-skill. PMC Open Access workflows belong in ncbi-pmc-skill. Datasets v2 workflows belong in ncbi-datasets-skill.
• Re-run requests in long conversations instead of relying on older tool output.
• Treat displayed ... in tool previews as UI truncation, not literal request content.

Execution behavior

• Return concise markdown summaries from the script output by default.
• In final user-facing summaries, never display a bare PMID or DOI. Render every PMID as a Markdown link in the form [PMID <PMID>](https://pubmed.ncbi.nlm.nih.gov/<PMID>/) and every DOI as [<DOI>](https://doi.org/<DOI>), including in tables, bullets, parentheticals, and source lists.
• Return raw JSON or XML only if the user explicitly asks for machine-readable output.
• Prefer targeted endpoint calls instead of broad unfiltered dumps.
• If the user needs the full raw response, set save_raw=true and report the saved file path.

Input

• Read one JSON object from stdin.
• Required field: endpoint
• Optional fields: params, record_path, response_format, max_items, max_depth, timeout_sec, save_raw, raw_output_path
• Common Entrez patterns:
  • {"endpoint":"esearch","params":{"db":"pubmed","term":"KRAS AND colorectal cancer","retmode":"json","retmax":10},"max_items":10}
  • {"endpoint":"esummary","params":{"db":"gene","id":"7157","retmode":"json"},"max_items":10}
  • {"endpoint":"efetch","params":{"db":"protein","id":"NP_000537.3","retmode":"xml"},"response_format":"xml","max_items":10}
  • {"endpoint":"elink","params":{"dbfrom":"gds","db":"pubmed","id":"200000001","retmode":"json"},"max_items":10}

Output

• Success returns ok, source, endpoint metadata, and either compact records, a compact summary, or text_head.
• Use raw_output_path when save_raw=true.
• Failure returns ok=false with error.code and error.message.

Execution

echo '{"endpoint":"esearch","params":{"db":"gene","term":"TP53[gene] AND human[orgn]","retmode":"json","retmax":10},"max_items":10}' | python scripts/ncbi_entrez.py

References

• Load references/geo.md only when the user specifically needs GEO query patterns.
• Keep the import package limited to this file, references/geo.md, and scripts/ncbi_entrez.py.