Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Struct Predictor · FreedomIntelligence | DecimalAI

Struct Predictor

SkillScore5Maintained

struct-predictor

Local protein structure prediction with AlphaFold, Boltz, or Chai. Compare predicted structures, compute RMSD, visualise 3D models.

scienceby FreedomIntelligence

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Research pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/struct-predictor/SKILL.md

decimalai skills pull struct-predictor --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/struct-predictor/SKILL.md https://app.decimal.ai/s/struct-predictor/SKILL.md

Raw file: /s/struct-predictor/SKILL.md ↗ · machine-readable summary at /s/struct-predictor.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 59d ago

Yes — +5% lift+1% tokens0% turns5% of whole cases pass with skill (1/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-22	✗→✓	▲ Improved	—	—
case-12	✗→✗	＝ Same ✗	—	—
case-10	✗→✗	＝ Same ✗	—	—
case-11	✗→✗	＝ Same ✗	—	—
case-14	✗→✗	＝ Same ✗	—	—

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 7/30/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

FreedomIntelligence / OpenClaw-Medical-Skills

skills/struct-predictor/SKILL.md

View on GitHub ↗

Imported 8/1/2026Version v1

Skill contents

SKILL.md

Struct Predictor

You are the Struct Predictor, a specialised agent for protein structure prediction and analysis.

Core Capabilities

Structure Prediction: Run AlphaFold (ColabFold), Boltz-1, or Chai locally
PDB Retrieval: Fetch experimental structures from PDB via OpenBio
Structure Comparison: Compute RMSD, TM-score between predicted and reference structures
Confidence Mapping: Visualise pLDDT and PAE confidence metrics
Report Generation: Markdown with 3D renders, confidence plots, and comparison tables

Dependencies

colabfold_batch or boltz or chai (at least one local predictor)
biopython (PDB parsing)
Optional: pymol (3D rendering), py3Dmol (interactive visualisation)

Example Queries

"Predict the structure of this protein sequence: MKWVTF..."
"Compare AlphaFold prediction of BRCA1 to the experimental PDB structure"
"Show the pLDDT confidence plot for my predicted structure"
"What is the RMSD between these two PDB files?"

Status

Planned -- implementation targeting Week 4-5 (Mar 20 - Apr 2).

Illustrative Want to see what struct-predictor would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

7/26/2026, 13:05 UTC · 22 cases

Lift

+5%

5% vs 0%

Agent turns

0%

1.0 vs 1.0

Tokens

+1%

2,178.27 vs 2,163.45

fail → pass · 1fail → fail · 21

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-12 If ColabFold is not available on our system, what alternative modern local end-to-end structure prediction tools should be used for monomer or complex modeling?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-10 Run MODELLER using a comparative homology modeling script with PIR alignment files to generate 10 loop-refined homology models.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-11 I want to run a local batch prediction on my FASTA file `targets.fasta`. Instead of using an online web server, what local command line tool executable associated with ColabFold should be invoked?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-14 Explain how to evaluate structural divergence between two PDB files using BioPython or PyMOL. Which specific numeric metric measures coordinate distance after optimal alignment?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-21 If the output report requires interactive 3D web viewing of the predicted PDB file inside a Jupyter notebook or HTML output, what Python library should be integrated?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-22 When setting up environment dependencies for structure prediction, list the three alternative local prediction engine options supported by the workflow.	fail→pass	—	—	—	—	—	—	—	—	—	—	—	—
case-05 What two standard confidence metrics should be extracted and visualized to assess per-residue local accuracy and relative domain positioning in a predicted structure?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-04 I have a predicted protein structure and an experimental reference structure. When comparing these two 3D models, what quantitative structural comparison metrics should be calculated to evaluate superimposition quality?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-03 We need to fetch experimental PDB structures directly in Python for structure validation. Rather than scraping raw HTML from the PDB website or making custom HTTP requests, what Python library ecosystem should we use to retrieve and parse PDB files?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-20 To produce static publication-quality 3D renders of a protein structure with pLDDT coloring for a markdown summary, what software tool can be scripted?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-09 Set up a GROMACS molecular dynamics simulation system for my protein model in an explicit SPC/E water box with 0.15 M NaCl and generate a topol.top file.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-08 I have a predicted protein target and a small molecule ligand SDF file. Please write an AutoDock Vina script to perform molecular docking and calculate binding free energy.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-16 When generating confidence visualisations for a predicted protein structure, how is pLDDT mapped across the residues?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-02 I want to set up an automated local prediction pipeline for protein sequences on my GPU server. Should I send REST requests to the ESMFold API, or what software should I install locally to run structure predictions?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-19 Which Python package is essential in this workflow for reading, parsing, and manipulating PDB coordinate files?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-06 What tools should be included in the environment to generate 3D structure renders and interactive visual widgets inside Markdown reports?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-17 Describe how PAE (Predicted Aligned Error) should be plotted and interpreted in the prediction analysis report.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-15 When comparing two protein folds that have similar global topology but flexible loop regions, why is TM-score calculated alongside RMSD?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-01 I need a structure prediction for the sequence MKTLLILAVVAAALAAGPAQA. Please generate the predicted model and return a report with pLDDT confidence charts and PAE heatmaps.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-13 Write a short Python code snippet showing how to programmatically download PDB file `1TUP` using standard OpenBio / BioPython PDB modules.	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-07 Design an output template for a protein structure prediction report. What specific visual and tabular sections should be included in the resulting Markdown document?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—
case-18 When summarizing structure comparisons between predicted and experimental PDB models, what tabular format should be used in the output markdown?	fail→fail	—	—	—	—	—	—	—	—	—	—	—	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

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. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 0%
Pass rate with the skill: 4.6%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 7/26/2026, 13:05 UTC

The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run5

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 58 days ago

Other measured skills in the registry, with their headline benchmark lift.