Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search and retrieve 3D protein structures from the RCSB Protein Data Bank
.claude/skills/brycewang-stanford-pdb-structure-api/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 145% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 182% | 0% |
| case-02 | ✓→✓ | = Same ✓ | 314% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 209% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 205% | 0% |
The RCSB Protein Data Bank (PDB) is the single global archive for experimentally determined 3D structures of biological macromolecules. It hosts over 200,000 structures resolved by X-ray crystallography, cryo-EM, NMR spectroscopy, and other methods. Each entry includes atomic coordinates, experimental metadata, polymer sequences, bound ligands, and literature references.
Two complementary APIs are available. The Data API (data.rcsb.org) serves structured entry metadata, polymer entities, and chemical components via RESTful GET endpoints. The Search API (search.rcsb.org) supports full-text, attribute-based, sequence similarity, and structure similarity searches.
No authentication required. Both APIs are freely accessible without API keys, tokens, or registration.
Retrieve metadata for a structure including experimental method, resolution, citations, and bound components.
GET https://data.rcsb.org/rest/v1/core/entry/{pdb_id}bashcurl "https://data.rcsb.org/rest/v1/core/entry/4HHB"
json{ "rcsb_id": "4HHB", "struct": { "title": "THE CRYSTAL STRUCTURE OF HUMAN DEOXYHAEMOGLOBIN AT 1.74 ANGSTROMS RESOLUTION" }, "exptl": [{"method": "X-RAY DIFFRACTION"}], "rcsb_entry_info": { "deposited_atom_count": 4779, "molecular_weight": 64.74, "polymer_composition": "heteromeric protein", "polymer_entity_count_protein": 2, "resolution_combined": [1.74], "nonpolymer_bound_components": ["HEM"] } }
Retrieve protein/nucleic acid entity details including sequence, organism, and gene info.
GET https://data.rcsb.org/rest/v1/core/polymer_entity/{pdb_id}/{entity_id}bashcurl "https://data.rcsb.org/rest/v1/core/polymer_entity/4HHB/1"
json{ "entity_poly": { "pdbx_seq_one_letter_code_can": "VLSPADKTNVKAAWGKVGAHAGEYGAEALERMFLSFPTTKTYFPHFDLSH...", "rcsb_entity_polymer_type": "Protein", "rcsb_sample_sequence_length": 141, "type": "polypeptide(L)" }, "entity_src_gen": [{ "gene_src_common_name": "Human", "pdbx_gene_src_scientific_name": "Homo sapiens", "pdbx_gene_src_ncbi_taxonomy_id": "9606" }] }
Retrieve ligand or small molecule metadata by component ID.
GET https://data.rcsb.org/rest/v1/core/chemcomp/{comp_id}bashcurl "https://data.rcsb.org/rest/v1/core/chemcomp/HEM"
json{ "rcsb_id": "HEM", "chem_comp": { "formula": "C34 H32 Fe N4 O4", "formula_weight": 616.487, "name": "PROTOPORPHYRIN IX CONTAINING FE", "type": "non-polymer" } }
Search across all PDB entries with free-text queries. Returns ranked results by relevance.
POST https://search.rcsb.org/rcsbsearch/v2/queryContent-Type: application/jsonquery.type ("terminal"), query.service ("full_text", "text", "sequence", "structure"), query.parameters.value, return_type ("entry", "polymer_entity", "assembly"), request_options.paginate.start/rowsbashcurl -X POST "https://search.rcsb.org/rcsbsearch/v2/query" \ -H "Content-Type: application/json" \ -d '{ "query": { "type": "terminal", "service": "full_text", "parameters": {"value": "hemoglobin"} }, "return_type": "entry", "request_options": { "results_content_type": ["experimental"], "paginate": {"start": 0, "rows": 3} } }'
json{ "query_id": "6f7192a6-d65b-4ff1-9d94-37b9600a8864", "result_type": "entry", "total_count": 8960, "result_set": [ {"identifier": "3GOU", "score": 1.0}, {"identifier": "6IHX", "score": 0.9995}, {"identifier": "2PGH", "score": 0.9985} ] }
For attribute-based searches, use "service": "text" with "attribute" and "operator" fields. Combine multiple criteria with "type": "group" and "logical_operator": "and".
No formal rate limits or rate-limit headers are published. RCSB recommends reasonable request rates. For bulk data, use FTP downloads at https://files.rcsb.org/pub/pdb/ or ftp://ftp.wwpdb.org/pub/pdb/ instead of iterative API calls.
pythonimport requests # Search for kinase inhibitor structures search_body = { "query": {"type": "terminal", "service": "full_text", "parameters": {"value": "tyrosine kinase inhibitor"}}, "return_type": "entry", "request_options": {"results_content_type": ["experimental"], "paginate": {"start": 0, "rows": 5}} } results = requests.post("https://search.rcsb.org/rcsbsearch/v2/query", json=search_body).json() print(f"Total hits: {results['total_count']}") # Retrieve metadata for each hit for hit in results["result_set"]: pdb_id = hit["identifier"] entry = requests.get( f"https://data.rcsb.org/rest/v1/core/entry/{pdb_id}").json() info = entry["rcsb_entry_info"] print(f"{pdb_id}: {entry['struct']['title'][:80]}") print(f" Resolution: {info.get('resolution_combined', ['N/A'])[0]} A, " f"Method: {info['experimental_method']}")
pythonimport requests pdb_id = "4HHB" entry = requests.get( f"https://data.rcsb.org/rest/v1/core/entry/{pdb_id}").json() for eid in range(1, entry["rcsb_entry_info"]["polymer_entity_count"] + 1): entity = requests.get( f"https://data.rcsb.org/rest/v1/core/polymer_entity/{pdb_id}/{eid}" ).json() poly = entity["entity_poly"] src = entity.get("rcsb_entity_source_organism", [{}])[0] print(f"Entity {eid}: {poly['rcsb_entity_polymer_type']} " f"({src.get('ncbi_scientific_name', 'N/A')})") print(f" {poly['rcsb_sample_sequence_length']} residues: " f"{poly['pdbx_seq_one_letter_code_can'][:50]}...")
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,949 | 31,813 | +191% | 1 | 1 | 0% | 2,004 | 4,903 | +145% | 0 | 0 | — |
case-02 | pass→pass | 4,181 | 2,662 | -36% | 1 | 1 | 0% | 648 | 2,681 | +314% | 0 | 0 | — |
case-03 | pass→pass | 4,974 | 3,486 | -30% | 1 | 1 | 0% | 935 | 2,887 | +209% | 0 | 0 | — |
case-04 | fail→pass | 5,639 | 2,597 | -54% | 1 | 1 | 0% | 941 | 2,649 | +182% | 0 | 0 | — |
case-05 | pass→pass | 4,913 | 2,499 | -49% | 1 | 1 | 0% | 873 | 2,661 | +205% | 0 | 0 | — |
case-06 | pass→pass | 6,902 | 3,524 | -49% | 1 | 1 | 0% | 1,148 | 2,819 | +146% | 0 | 0 | — |
case-07 | pass→pass | 4,660 | 3,372 | -28% | 1 | 1 | 0% | 861 | 2,775 | +222% | 0 | 0 | — |
case-08 | pass→pass | 11,887 | 8,753 | -26% | 1 | 1 | 0% | 2,342 | 4,063 | +73% | 0 | 0 | — |
case-09 | pass→pass | 17,275 | 12,283 | -29% | 1 | 1 | 0% | 2,626 | 4,986 | +90% | 0 | 0 | — |
case-10 | fail→fail | 10,863 | 5,135 | -53% | 1 | 1 | 0% | 1,828 | 3,178 | +74% | 0 | 0 | — |
case-11 | pass→pass | 11,642 | 5,500 | -53% | 1 | 1 | 0% | 2,058 | 3,203 | +56% | 0 | 0 | — |
case-12 | pass→pass | 9,359 | 2,861 | -69% | 1 | 1 | 0% | 1,766 | 2,772 | +57% | 0 | 0 | — |
case-13 | pass→pass | 8,693 | 3,850 | -56% | 1 | 1 | 0% | 1,602 | 2,881 | +80% | 0 | 0 | — |
case-14 | pass→pass | 5,374 | 2,341 | -56% | 1 | 1 | 0% | 931 | 2,624 | +182% | 0 | 0 | — |
case-15 | pass→pass | 10,217 | 7,640 | -25% | 1 | 1 | 0% | 1,531 | 3,806 | +149% | 0 | 0 | — |
case-16 | pass→pass | 6,071 | 4,612 | -24% | 1 | 1 | 0% | 1,031 | 2,974 | +188% | 0 | 0 | — |
case-17 | pass→pass | 7,170 | 4,216 | -41% | 1 | 1 | 0% | 1,398 | 3,041 | +118% | 0 | 0 | — |
case-18 | pass→pass | 4,609 | 3,576 | -22% | 1 | 1 | 0% | 739 | 2,892 | +291% | 0 | 0 | — |
case-19 | pass→pass | 8,134 | 5,578 | -31% | 1 | 1 | 0% | 1,648 | 3,432 | +108% | 0 | 0 | — |
case-20 | pass→pass | 5,461 | 3,317 | -39% | 1 | 1 | 0% | 1,055 | 2,837 | +169% | 0 | 0 | — |
case-21 | pass→pass | 7,813 | 5,671 | -27% | 1 | 1 | 0% | 1,492 | 3,283 | +120% | 0 | 0 | — |
case-22 | pass→pass | 7,270 | 6,868 | -6% | 1 | 1 | 0% | 1,500 | 3,523 | +135% | 0 | 0 | — |
case-23 | pass→pass | 7,588 | 7,808 | +3% | 1 | 1 | 0% | 1,355 | 3,588 | +165% | 0 | 0 | — |
case-24 | pass→pass | 8,330 | 12,011 | +44% | 1 | 1 | 0% | 1,656 | 4,599 | +178% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted. The headline lift of +8 percentage points is the difference between those two pass rates over the 24 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.