What problem does it solve? Tuning an inverted pendulum simulation rig by hand is guesswork: gravity, rod length, push authority, and gusts interact in ways that are hard to predict, and an invalid configuration file silently breaks the simulation. This Skill guides the design of pendulum.json rigs and verifies them so the balancing behavior is dramatic but fair. ## Core Features & Use Cases - Rig Design: Edit pendulum.json parameters (gravity, length, maxTorque, gustEvery/gustStrength, style line) and watch Jev adapt live in the viewer without restarting. - Rig Validation: Run node "$JEV_DSH/toolchain/check.mjs" to reject invalid configs (missing title, out-of-range values like gravity 0.5..30, length 0.2..3, stepMs 30..2000) and print the unrecoverable lean angle. - Direct Model Queries: Use toolchain/jev.mjs to ask Jev which push action steadies a given rod state before committing to a rig, with a deterministic mock when TYPESAFE_API_KEY is unset. - Use Case: Tune gravity from 5 to 10, add gusts, and confirm the rod visibly struggles and occasionally drops rather than standing like a statue or falling instantly. ## Quick Start Ask the agent to update pendulum.json with gravity 7 and moderate gusts, then run the check script and watch the viewer to confirm the rod wobbles but recovers.
