Skill UI

Prove data/AI workflows end-to-end on tiny samples, find cost-saving wins, then scale within budget using cost-gated steps. Ideal for multi-stage pipelines involving paid compute like LLM tokens or warehouse credits.

Ten rules that keep coding agents from re-sending bloated context on every call: locate before reading, compute instead of dumping data, bound noisy command output, never re-read, batch parallel calls and answer lean. Cuts token cost ~32% with unchanged pass rates.

Drive native desktop apps through an Automation REPL: bind applications, inspect accessibility trees, type text, click, and capture screenshots while honoring OS permissions and user stop requests.

Runs Cantera homogeneous ideal-gas reactors for constant-volume or constant-pressure ignition, computes temperature-based ignition delay, checks conservation, and refines numerical settings with mechanism provenance.