SimFoundry AGENTS.md

A set of instructions for coding agents working on SimFoundry, a system that turns short real-world videos into physics-ready simulated scenes. It explains the project stages, required environment, safe files to edit, and saved-scene rules.

In plain words
What is it for?
Use it when modifying or reviewing SimFoundry, especially its reconstruction pipeline, scene editing workflow, generated tasks, or simulation outputs.
Why use it?
It prevents agents from missing project-specific assumptions or making changes that silently break later reconstruction, augmentation, or application stages.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/nvlabs/simfoundry/agents-md
Clone the repo
git clone --depth 1 https://github.com/NVlabs/SimFoundry

Made for: Codex, OpenCode.

Per session 4,224 This file is loaded in full into every session.
When invoked 4,224 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.04224 $0.04224
Opus 5 $0.02112 $0.02112
Sonnet 5 $0.00845 $0.00845
Haiku 4.5 $0.00422 $0.00422

Measured 2d ago against content hash 3e876c469e0f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

SimFoundry AGENTS.md 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.

AGENTS.md · 307 lines

How it starts

The opening of the file, as written. The whole thing — 307 lines — stays where its author put it; the contents beside it link to each section on GitHub.

SimFoundry — agent guide

Orientation for coding agents. This covers what the code does not say out loud: which environment to run in, which trees are safe to edit, and which contracts break silently. For what the project is, read README.md.

What this repo does

SimFoundry turns a short real-world video into a physics-ready OmniGibson scene: segment the objects, reconstruct geometry, generate textured meshes, compile a scene with physical parameters and task proposals. The pipeline runs in three stages under scripts/pipeline/:

Stage Directory Produces
A — reconstruction A_reconstruction/stages/ Depth, segmentation, meshes, s11_sim/scene_objects_info.json, s12_physics/pb_scene_poses.json, an OmniGibson scene
B — augmentation B_augmentation/stages/ Digital-cousin variants, scene sampling, task proposals
C — application C_application/stages/ Teleop, demo generation, policy evaluation

Scene editing sits between B and C: a human corrects what the pipeline got wrong — object poses, scales, cameras — before the scene is used for learning or evaluation. Everything downstream inherits those corrections.

Environments

There are many conda envs on a dev box (3dgrut, da3, hunyuan, …); most belong to a single pipeline stage. For general work:

  • simfoundry — the main env. OmniGibson, Isaac Sim, torch. Used by the pipeline, the OmniGibson editor, and settle.py. find_settle_python in light_editor/server.py checks that the env's omnigibson resolves inside this repo.
  • simfoundry-editor — the light editor's env, no GPU. Created by scripts/installation/install_light_editor.sh; re-run that after a pull. See the trap below.

Trap: omnigibson can resolve outside this checkout. An env may carry this repo's simfoundry package but a different clone's OmniGibson. The editor then runs against the wrong deps/ tree and stamps the wrong OmniGibson version into saved scenes. run_editor.sh warns about this at startup; do not ignore it. Check with python -c 'import omnigibson; print(omnigibson.__file__)'.

Read the full file on GitHub · 307 lines

Changes

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.

  1. 2d ago First seen · 307 lines · 4,224 tokens per session scan A 3e876c469e0f

Subscribe to this mod's changes

SimFoundry AGENTS.md is an instructions file published in the GitHub repository NVlabs/SimFoundry (290 stars, last pushed 6d ago), licensed Apache-2.0. It adds 4,224 tokens to every session, about $0.0211 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 2026-08-30.