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 adaptyvbio/protein-design-skills --skill mosaicgit clone --depth 1 https://github.com/adaptyvbio/protein-design-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/adaptyvbio/protein-design-skills/mosaic)<a href="https://agentmods.dev/skills/adaptyvbio/protein-design-skills/mosaic"><img src="https://agentmods.dev/badge/skills/adaptyvbio/protein-design-skills/mosaic/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/adaptyvbio/protein-design-skills/mosaic"><img src="https://agentmods.dev/badge/skills/adaptyvbio/protein-design-skills/mosaic.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 | $0.00136 | $0.01666 |
| Opus 5 | $0.00068 | $0.00833 |
| Sonnet 5 | $0.00027 | $0.00333 |
| Haiku 4.5 | $0.00014 | $0.00167 |
Grade A, and why
mosaic 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 12d 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 — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Mosaic Multi-Objective Design
Mosaic (Escalante Bio) is a JAX framework for "functional, multi-objective protein design using continuous relaxation." It optimizes a soft sequence by gradient descent over a continuous relaxation of sequence space, in the lineage of ColabDesign, RSO, and BindCraft, with one key difference: it composes multiple learned objectives from different models in a single differentiable loss.
When Mosaic fits
Mosaic is a framework for custom objectives, not a one-click method. The README is
explicit: it "may require substantial hand-holding (tuning learning rates, etc),
often produces proteins that fail simple in-silico tests, [and] should be combined
with standard filtering methods." Reach for it when a fixed pipeline cannot express
the objective you need. For a turnkey binder run, use bindcraft instead.
Prerequisites
| Requirement | Minimum | Recommended |
|---|---|---|
| Python | 3.11+ | 3.11 |
| Framework | JAX with CUDA or TPU | JAX CUDA 12 |
| GPU VRAM | 24GB | 48GB+ (depends on predictors used) |
JIT compilation makes the first call to any loss slow; later calls are fast.
Install
Mosaic runs locally on a JAX GPU or TPU build. It has no CLI and no Modal integration; you drive it through the marimo notebooks or the Python API.
git clone https://github.com/escalante-bio/mosaic && cd mosaic
uv sync --group jax-cuda # or --group jax-tpu / --group jax-cpu
uv add jax[cuda12] # may be needed for a GPU build
uv run marimo edit examples/example_notebook.py
Ready-made examples include esmfold_minibinder.py, esmfold_vhh.py,
boltzgen_pipeline.py, and batched_protenix.py.
Core idea
A design objective is built from LossTerm objects that you add and scale with plain
Python arithmetic, then hand to an optimizer.
import mosaic.losses.structure_prediction as sp
# Compose a loss from interface, confidence, and inverse-folding terms
design_loss = (
sp.BinderTargetContact()
+ sp.WithinBinderContact()
+ 0.05 * sp.TargetBinderPAE()
+ 0.05 * sp.BinderTargetPAE()
+ 0.025 * sp.IPTMLoss()
+ 0.1 * sp.PLDDTLoss()
)
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.
- 12d ago First seen · 170 lines · 136 tokens per session scan A a67d371072bf
mosaic is a skill published in the GitHub repository adaptyvbio/protein-design-skills (158 stars, last pushed 3mo ago), licensed MIT. It adds 136 tokens to every session and 1,666 once invoked, about $0.0007 per session on Opus 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
xxd-color-brief
Turn vague visual direction into a practical Chinese traditional color brief. Use when a user has mood words, client feedback, audience positioning, references, or an unclear Chinese traditional color direction that must become palette constraints before choosing colors.
xxd-ui-token
Convert Chinese traditional colors into practical UI design tokens. Use when a user asks for CSS variables, Tailwind theme colors, Figma variables, light and dark modes, semantic color roles, component states, or designer developer handoff based on Chinese traditional colors.
xxd-accessible-color
Check and repair Chinese traditional color combinations for readability and accessibility. Use when a user asks whether text, buttons, UI states, charts, light mode, dark mode, or traditional color palettes meet contrast and non-color-cue requirements.
xxd-brand-system
Build a reusable brand color system from Chinese traditional colors. Use when a user needs brand palette governance, color rules, identity consistency, usage ratios, forbidden combinations, cross-channel guidance, or long-term visual rules based on Chinese traditional colors.
xxd-content-series
Design reusable Chinese traditional color systems for content series. Use when a user needs color rules for Xiaohongshu covers, WeChat articles, video thumbnails, course slides, newsletters, recurring columns, editorial series, carousels, or multi-part content that must stay recognizable over time.
xxd-data-viz
Create chart and data visualization palettes from Chinese traditional colors. Use when a user needs categorical, sequential, diverging, highlight, dashboard, map, ECharts, D3, Chart.js, or colorblind-aware data palettes with Chinese traditional color identity.