Skills
Featured scientific & agent skills, skills.sh catalog, and 1-click Skill crystallization.
Skills are packaged SKILL.md instructions, reference ontologies, and scripts that agents load on demand. Only a compact index is injected at session start (keeping the prompt budget small), while the full SKILL.md body loads progressively when invoked via /skill-name or matched to a task.
Featured Skills
Biomolecular structure prediction, protein–ligand complexes, and binding affinity estimation with Boltz-1 / Boltz-2.
De novo protein binder design, sequence generation, structure filtering, and candidate ranking.
Monomer and multimer protein folding with pLDDT, PAE confidence matrices, and interactive 3D structure output.
Coarse-grained molecular dynamics simulations for intrinsically disordered proteins (IDPs) and phase separation.
All-atom protein–ligand molecular dynamics setup, equilibration, trajectory analysis, and water-network profiling.
Model entities, Supports/Contradicts claims, and Git-style review/* branches across the 6 Omnigraph Cookbook modes (01–06).
Curate USER.md & MEMORY.md snapshots, write Living Wiki [[wikilinks]], and crystallize domain knowledge into reusable skills.
Develop, benchmark, and scale GPU experiments and parallel sub-agent workflows on Modal containers.
Convert PDFs, Word (.docx), PowerPoint (.pptx), Excel (.xlsx), and scientific manuscripts into clean Markdown.
Author signaling pathways, architecture flowcharts, sequence diagrams, and state machines in Mermaid.
Synthesize experimental runs, figures, and executive takeaways and post structured updates to Slack.
Build interactive static web dashboards from session results, preview locally, and publish to sandwalk.site.
Remote SLURM GPU cluster execution over SSH with sbatch/srun, Pyxis/Enroot containers, and shared Lustre storage.
Schedule recurring agent workflows, one-shot follow-ups, thread snoozing, and webhook triggers.
Manage SSH servers, execute remote commands, sync files via rsync, and auto-bootstrap persistent remote Sandwalk daemons.
Curated Skill Collections (One-Command Install)
Install entire collections of scientific and engineering skills from GitHub or Dashboard → Skills:
| Provider | Install Command | GitHub Repository |
|---|---|---|
| FastFold AI | sandwalk skills add fastfold-ai/skills | fastfold-ai/skills |
| K-Dense AI | sandwalk skills add K-Dense-AI/scientific-agent-skills | K-Dense-AI/scientific-agent-skills |
| Anthropic Life Sciences | sandwalk skills add anthropics/life-sciences | anthropics/life-sciences |
| Google DeepMind | sandwalk skills add google-deepmind/science-skills | google-deepmind/science-skills |
| Hugging Face | sandwalk skills add huggingface/skills | huggingface/skills |
| NVIDIA Skills | sandwalk skills add NVIDIA/skills@skills/tao-run-on-slurm | NVIDIA/skills |
| Firecrawl Anydoc | sandwalk skills add firecrawl/anydoc | firecrawl/anydoc |
Crystallize Your Own Skills (Create Skill / memory.create_skill)
You can also generate new skills directly from what your workspace has learned (~/.sandwalk/skills/<slug>/SKILL.md + references/{schema.pg,queries.gq,wiki-digest.md}):
- From the UI: In Dashboard → Memory & Wiki, click Create Skill in the top header (or To Skill inside any Living Wiki page or Knowledge Graph view).
- From Chat or Capy: Ask any agent "Create a skill from what we know about oral GLP-1R" (calls
memory.create_skillorcapy.create_skill_from_memory).
Using Skills in the Composer
Type / in the composer (or open + → Skills) to attach any installed or crystallized skill to your message. See also CLI Skills.
Last updated on