Identity & context
Stable IDs, versions, units, coordinate frames, and time stay with the values they qualify.
Grant Pedersen · Product designer · New York
Science UX
An AI skill by Grant Pedersen for scientific software, preserving identity, units, provenance, uncertainty, and reproducible workflows.
AI Skills · Scientific software · 2026
Keep the science attached to the interface.
Scientific tools can become easier to use while keeping the detail that makes a result interpretable. This skill gives an agent a way to work with that tension across research workspaces, instruments, simulations, analysis, and visualization.
Version 1.1.0 improves how agents recognize simulation-results reviews and adds recorded evaluations using a synthetic results screen. The review follows inputs, resolved settings, output lineage, and failed-run recovery while keeping computational completion separate from scientific validity.
What the interface needs to preserve.
Provenance & comparison
Inputs, settings, transformations, and output lineage remain inspectable. A rerun becomes a linked new run.
Evidence & uncertainty
Observed, derived, simulated, and AI-suggested information stay distinct. Missing data never quietly becomes zero.
Control & recovery
Requested, running, partial, failed, and uncertain outcomes have different meanings and recovery paths.
A results screen, beyond the chart.
The review follows the selected run back to its inputs and resolved settings, checks whether comparisons retain their context, and examines what the user can safely do after a failure.
Review a simulation results screen for run comparison, reproducibility, and recovery from a failed run. Preserve its existing visual language and report only findings supported by the interface and code.
Guidance close to the work.
- Workflows & scientific state
Objects, instruments, records, computational jobs, and recovery.
- Visualization & interaction
Scientific views, uncertainty, selection, and accessible interaction.
- Evidence & evaluation
Research claims, review priorities, and representative user testing.
The skill supports interface design and review. Scientific conclusions and product-specific experimental rules remain with the appropriate domain experts.
Use the skill.
npx skills add grant-atl/science-healthcare-ux --skill scientific-software-ux -a codex -a claude-code