DART is an open-source C++23 physics engine that simulates the movement and interactions of articulated rigid-body systems for robotics, animation, and machine learning. Researchers and developers use it for kinematics, dynamics, collision handling, constraints, and loading robot models, with C++ and Python interfaces. The catalogue add-ons support workflows built around this engine.
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 dartsim/dart --skill dart-verify-simgit clone --depth 1 https://github.com/dartsim/dartWrote 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/dartsim/dart/dart-verify-sim)<a href="https://agentmods.dev/skills/dartsim/dart/dart-verify-sim"><img src="https://agentmods.dev/badge/skills/dartsim/dart/dart-verify-sim/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/dartsim/dart/dart-verify-sim"><img src="https://agentmods.dev/badge/skills/dartsim/dart/dart-verify-sim.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Agent Snooping · line 7 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
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 | $0.00032 | $0.01674 |
| Opus 5.5 | $0.00013 | $0.00670 |
| Sonnet 5.5 | $0.00006 | $0.00335 |
| Haiku 4.5 | $0.00003 | $0.00167 |
Grade A, and why
dart-verify-sim 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 4d 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 — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DART Simulation Verification
Load this skill whenever a DART claim depends on 3D structure or behavior: model/scene loading, dynamics, collision/contact/constraints, simulation stepping, Filament rendering, or a visual example.
Lead with text, corroborate with images. An image-capable agent can inspect a rendered frame, but pixels do not expose solver state and machine pixel checks are not semantic inspection. First run a text oracle, then verify the same claim end to end through the core Filament capture/debug path.
Owner And Evidence Contract
Read docs/ai/verification.md.
For Python catalog scene changes, also follow
docs/onboarding/testing.md § "Demo Catalog Verification" for construction,
scripted controls, and matched-capture parity evidence.
Record:
- a text correctness oracle (focused behavioral test, poses/velocities,
contact trace, or
pixi run bm-boxes-headlesschecksum comparison); - an assessed headless capture with only the debug layers needed by the claim; and
- an actual semantic inspection of the selected still or temporal frames; and
- the reproduce command, view-quality result, reconciliation, verdict, claim boundary, and limitations.
If Filament capture is unavailable or visual corroboration is genuinely irrelevant,
state why and name the replacement evidence. Never use an image as the
sole correctness oracle.
Image-capable Review Loop
Every model lane routed in docs/ai/README.md § "Updating Models And Coding
Agents" supports native image input and original-detail inspection. Keep this
loop capability-based; update model facts in that owner doc, not here.
- State one claim, its expected visible observation, and the text oracle that decides correctness.
- Capture one assessed, claim-tied view first. Add only the debug layers needed by the claim; use paired plain/debug views when overlays obscure geometry. For turntable or motion output, inspect at least the sidecar's start/middle/end targets and add intervening frames or a sheet when a transient event matters.
- Run
image-verdictfor artifact integrity, then open the selected PNG with the active agent's native image viewer. Use original detail for fine contacts, labels, bounds, or frame axes. A passing view report orimage-verdictis not semantic visual review. - Record the visible observation separately from the text result. If they disagree, do not average them into a pass: report fail or uncertain, inspect the capture sidecar, reframe or recapture, and investigate simulation state.
- Close with pass/fail/uncertain, artifact path, camera/layers, reproduce
command, what the image shows, and what it does not prove. If native image
review is unavailable, hand the selected files and sidecar to an
image-capable reviewer with
pixi run verification-bundle -- ...and record that limitation.
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.
- 4d ago Changed · +9 lines bde638309f26
- 16d ago Changed · +8 lines b7da1806b58b
- 17d ago Changed · +5 lines · -2 tokens per session e5ed34e04b6f
- 19d ago Changed 933fe68cd7ab
- 20d ago Changed · -7 lines f672763de7d1
- 22d ago Changed · +11 lines bfe0cd5be0a2
- 24d ago Changed · -2 lines d8b32fb79b5b
- 1mo ago Changed · -14 lines · -6 tokens per session 8fac5a0c2b11
- 1mo ago Changed · -1 lines 5e3cfc32cade
- 1mo ago First seen · 137 lines · 40 tokens per session scan A 55d4deb0d3b4
dart-verify-sim is a skill published in the GitHub repository dartsim/dart (1,222 stars, last pushed yesterday), licensed BSD-2-Clause. It adds 32 tokens to every session and 1,674 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-09-01.
Other skills, from other repositories
sdf
SDFormat/SDF model and world authoring, validation, and simulator handoff. Use for .sdf files, SDFormat XML, models, worlds, links, joints, poses, frames, inertials, visual/collision geometry, mesh URIs, sensors, lights, physics, plugins, includes, Gazebo, static SDF review, or simulator-specific metadata. Do not use…
regression-suite
Map test coverage to GDD critical paths, find fixed bugs lacking regression tests, flag drift from new features.
soak-test
Soak test protocol for extended play — what to observe and log for slow leaks, fatigue, late-appearing edge cases.
team-polish
Orchestrate the polish team — performance-analyst, technical-artist, sound-designer, qa-tester — to optimize and harden a feature or area.
evaluating-cosmos-policy
Evaluates NVIDIA Cosmos Policy on LIBERO and RoboCasa simulation environments. Use when setting up cosmos-policy for robot manipulation evaluation, running headless GPU evaluations with EGL rendering, or profiling inference latency on cluster or local GPU machines.
smoke-check
Critical-path smoke gate before QA hand-off — runs the automated suite. A failed check means the build is not QA-ready.