Science Skills is a collection of add-ons that give AI agents structured instructions, scripts, and references for scientific research, including genomics, structural biology, cheminformatics, and literature search. Researchers use it to guide agents through specialized scientific tasks with information from databases and tools such as AlphaGenome, AFDB, and UniProt. The catalogue entries are individual skills from this collection.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add google-deepmind/science-skills --skill foldseek_structural_searchgit clone --depth 1 https://github.com/google-deepmind/science-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/google-deepmind/science-skills/foldseek_structural_search)<a href="https://agentmods.dev/skills/google-deepmind/science-skills/foldseek_structural_search"><img src="https://agentmods.dev/badge/skills/google-deepmind/science-skills/foldseek_structural_search/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/google-deepmind/science-skills/foldseek_structural_search"><img src="https://agentmods.dev/badge/skills/google-deepmind/science-skills/foldseek_structural_search.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00097 | $0.01286 |
| Opus 5.5 | $0.00039 | $0.00514 |
| Sonnet 5.5 | $0.00019 | $0.00257 |
| Haiku 4.5 | $0.00010 | $0.00129 |
Grade A, and why
foldseek-structural-search scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Prerequisites
uv: Read theuvskill and follow its Setup instructions to ensureuvis installed and on PATH.- User Notification: If .licenses/foldseek_structural_search_LICENSE.txt does not already exist in the workspace root directory then (1) prominently notify the user to check the terms at https://search.foldseek.com/search and https://github.com/steineggerlab/foldseek, then (2) create the file recording the notification text and timestamp.
Goal
Submit a user-provided 3D protein structure file (.cif, .mmcif, or .pdb)
to the Foldseek web server API to find structurally similar proteins. Report the
top structural hits, interpret key alignment metrics, summarize the inferred
protein functions, save the Markdown-formatted table to a .md file, and save
the full detailed results to a local JSON file.
Core Rules
- File Requirement: This tool absolutely cannot search by sequence, name,
or accession ID. It strictly requires a
.pdb,.cif, or.mmciffile path. - Strict Validation: Never bypass the input validation or the database allowlist check.
- Do Not Parse the JSON: Rely entirely on the generated
.mdfile for your immediate summary. The JSON is saved purely for subsequent, specialized tool use. - No Raw Parsing: Do not attempt to parse or read the raw 3D coordinates yourself; always pass the file to the script.
- Notification: If this skill is used, ensure this is mentioned in the output.
Instructions
- Strict Input Validation: Verify that the user has explicitly provided a
valid path to a
.cif,.mmcif, or.pdbfile in their workspace.- If the user provided a protein name, an amino acid sequence, or an accession ID (e.g., a UniProt ID) but NO downloaded structure file, halt immediately. Do not run the script.
- Inform the user that Foldseek requires a physical 3D coordinate file, and suggest downloading the structure first (e.g., using the AlphaFold fetch tool).
- Database Validation: Check if the user requested specific databases to
search.
- Allowed List:
afdb50,afdb-swissprot,pdb100,BFVD,mgnify_esm30,cath50,gmgcl_id,bfmd,afdb-proteome. - If the user requests a database NOT on this list, halt immediately. Do not run the script. Inform the user that the database is unsupported and provide them with the allowed list.
- Allowed List:
- Generate File Names: Generate descriptive output file names for both the
JSON data and the Markdown table based on the input file (e.g.,
proteinA_foldseek_results.jsonandproteinA_foldseek_results.md). - Execute the python script based on the user's request, redirecting the
standard output into your generated
.mdfile:- Default (No databases specified):
uv run scripts/search.py <path-to-file> -o <generated-filename.json> > <generated-filename.md> - Custom (Valid databases specified):
uv run scripts/search.py <path-to-file> -o <generated-filename.json> --databases <db1,db2,db3> > <generated-filename.md>
- Default (No databases specified):
- The script will query the databases, save the full JSON payload, and write a
Markdown-formatted table to your specified
.mdfile. - Read the Results: Open and read the newly generated
.mdfile carefully to view the Markdown table. - Interpret the Metrics: Summarize the top 3 to 5 structural matches that
have meaningfull annotations for the user. When reporting, assess the match
quality using these specific fields:
- Prob (Probability): Values approaching 1.0 (100%) indicate extreme confidence that the fold is a true structural homologue.
- Q-Cov (Query Coverage): High percentages mean the match covers the majority of the query protein's overall shape, rather than just a small local motif.
- E-value & Seq Identity: Use these to provide additional evolutionary context.
- Perform Functional Analysis: Analyze the text descriptions embedded
within the
Target IDcolumn for the reported matches.- Explicitly report the specific protein names/functions of the top structural homologues.
- Provide a synthesized overview summarizing the entire variety of different functions, domains, or protein families found across the whole list of homologues (e.g., "Most hits are portal proteins, but there is also a distinct cluster of viral capsid matches...").
- Explicitly inform the user of both newly created files (
.jsonand.md) and their locations so they can be seamlessly used in subsequent analysis steps.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 100 lines · 97 tokens per session scan A 3aff6e5f4725
foldseek-structural-search is a skill published in the GitHub repository google-deepmind/science-skills (3,307 stars, last pushed 6d ago), licensed Apache-2.0. It adds 97 tokens to every session and 1,286 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
instrument-data-to-allotrope
Convert laboratory instrument output files (PDF, CSV, Excel, TXT) to Allotrope Simple Model (ASM) JSON format or flattened 2D CSV. Use this skill when scientists need to standardize instrument data for LIMS systems, data lakes, or downstream analysis. Supports auto-detection of instrument types. Outputs include full…
exploratory-data-analysis
Performs bounded, local exploratory analysis of explicitly supported scientific files. Supports redacted CSV/TSV/JSON profiles; optional NumPy, HDF5, FASTA/FASTQ, and basic image metadata inspection; missingness/leakage audits; outlier and transformation sensitivity; and rigorous EDA report scaffolds. Other domain…
matlab
Builds, reviews, migrates, and plans MATLAB or GNU Octave numerical workflows. Use for arrays, tabular/time data, tests, projects, graphics, MAT files, and explicit Python interoperability.
phylogenetics
Builds and analyzes phylogenetic trees using MAFFT multiple sequence alignment, IQ-TREE maximum likelihood with ModelFinder and branch support, and FastTree approximate inference. Uses ETE3 for tree summaries and visualization. Applies to homologous nucleotide or protein sequences, microbial gene trees, protein…
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
presenting-conference-talks
Generates conference presentation slides (Beamer LaTeX PDF and editable PPTX) from a compiled paper with speaker notes and talk script. Use when preparing oral talks, spotlight presentations, or invited talks for ML and systems conferences.