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.
Borrowing it
Nothing to install: this file belongs to dartsim/dart. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dartsim/dart/main/.agents/skills/dart-retro/SKILL.mdgit 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-retro)<a href="https://agentmods.dev/skills/dartsim/dart/dart-retro"><img src="https://agentmods.dev/badge/skills/dartsim/dart/dart-retro/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-retro"><img src="https://agentmods.dev/badge/skills/dartsim/dart/dart-retro.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 | $0.00019 | $0.00830 |
| Opus 5 | $0.00010 | $0.00415 |
| Sonnet 5 | $0.00004 | $0.00166 |
| Haiku 4.5 | $0.00002 | $0.00083 |
Grade A, and why
dart-retro 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 2d 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 — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
dart-retro
Use this skill in Codex to run the DART dart-retro workflow. The editable
workflow source lives in .claude/commands/; this file is its generated adapter
in the shared .agents/skills/ catalog.
Invocation
- Claude Code:
/dart-retro <arguments> - Codex:
$dart-retro <arguments>
Treat the text after the skill name as $ARGUMENTS. When the workflow
references $1, $2, etc., map those to the positional values supplied by the
user.
Command Body
Improve future task execution through a retrospective: $ARGUMENTS
Use the original request and session evidence to improve the harness for future similar tasks: better quality, completeness, or efficiency. Successful, unsatisfactory, incomplete, and blocked outcomes qualify. PR merge and dev-task retirement do not alone establish that the user's goal was met.
Required Reading
@AGENTS.md @docs/AGENTS.md @docs/ai/principles.md @docs/ai/components.md @docs/ai/verification.md
Load task owners as needed. Read docs/onboarding/ai-tools.md for tool/runtime
questions and before controlled agent runs. docs/ai/components.md owns
placement of accepted learnings; this workflow owns the retrospective method.
Skip If
Conclude without harness edits when the evidence supports no reusable improvement: existing guidance already covers the lesson and is discoverable, the finding is task-specific, or no change has a supported benefit. State why and report missing evidence; success, failure, or a merged PR alone is not a skip condition.
Workflow
- Reconstruct intent and outcome. Default to the current session. Compare the initial request and later scope decisions with artifacts and results. Inspect relevant history around decisions, corrections, and failures, including substantive domain work before CI/review/closeout. State missing evidence; do not load every log by default.
- Find the harness contribution. Connect useful decisions, rework, missed requirements, and wasted context/tool cycles to instructions, routing, tools, or gates. Separate observations from inferred causes, implementation bugs, and external blockers. Check existing owners and executable coverage before proposing rules; investigate why existing guidance was missed.
- Choose a testable improvement. State the observed decision, causal gap, owner, and what an agent starting from the same brief should do differently. Name the expected benefit, a check that could disprove it, and successful constraints to preserve.
- Improve existing owners. Prefer removal, consolidation, or rewriting to
appending rules or files; repair discovery when guidance already exists.
Follow
docs/ai/components.mdfor placement and adapter regeneration. Keep session identifiers out of durable guidance. Preserve model/effort/action limits; a retrospective does not reopen the original implementation or authorize new GitHub mutations. - Validate proportionately. For consequential instruction changes, replay
the observed decision and a contrasting similar task; use fresh controlled
agents when permitted and useful. Distinguish structural checks, predicted
benefits, and measured outcomes. Run the relevant gates in
docs/ai/verification.mdandpixi run lintbefore committing.
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.
- 2d ago Changed 96e811fbec95
- 4d ago Changed · +7 lines · +4 tokens per session 61a012ced696
- 6d ago Changed · -3 lines 2b4221f4d347
- 8d ago First seen · 84 lines · 15 tokens per session scan A cfe935e1b9dc
dart-retro is a skill published in the GitHub repository dartsim/dart (1,204 stars, last pushed 2d ago), licensed BSD-2-Clause. It adds 19 tokens to every session and 830 once invoked, about $0.0001 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 2026-09-01.
Other skills, from other repositories
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.
cad-viewer
Start CAD Viewer and return review links for CAD and robot-description files. Use when visually reviewing .step, .stp, .glb, .stl, .3mf, .dxf, .urdf, .srdf, or .sdf files, especially when handed off from CAD, URDF, SRDF, or SDF generation skills.
urdf
URDF robot description authoring and validation. Use when creating, editing, inspecting, validating, or debugging .urdf files, robot links, joints, limits, inertials, visual/collision geometry, mesh references, frame conventions, or robot-description artifacts. Use the SRDF skill for MoveIt2 semantic groups and…
dfam-check
Measure mesh files against Design for Additive Manufacturing (DfAM) rules and report printability findings per process (FDM, SLS, SLA/DLP, metal PBF, MJF). Use when the user asks whether a part is printable, wants overhang/wall-thickness/support analysis of an .stl, .obj, .ply, or .3mf mesh, wants a build-orientation…
step-parts
Find, evaluate, and download common purchasable CAD parts from step.parts, including named off-the-shelf actuators, servos, motors, electronics boards, connectors, screws, bolts, nuts, washers, bearings, standoffs, and other catalog components. Use when Codex needs to search the hosted step.parts catalog before…
fine-tuning-serving-openpi
Fine-tune and serve Physical Intelligence OpenPI models (pi0, pi0-fast, pi0.5) using JAX or PyTorch backends for robot policy inference across ALOHA, DROID, and LIBERO environments. Use when adapting pi0 models to custom datasets, converting JAX checkpoints to PyTorch, running policy inference servers, or debugging…