NVIDIAKnown publisherNVIDIA/NuRec/skills/nurec-onboard/AGENTS.md

NuRec / nurec-onboard

AGENTS.md33 starsChanged 16 days agoApache-2.0View on GitHub

Project-level conventions for autonomous workflow execution. - Always assign container names. - Preserve the resource, GPU, environment, user, and container-side mount flags in the workflow's Docker templates. Substitute only documented placeholders. - Issue each Docker command as a single docker invocation built from literal words. Docker runs on the host rather than in the command sandbox, and that applies only to a command in this shape. A chained, piped, redirected, wildcard-bearing, environment-assignment-prefixed, $(...)-bearing, or $VAR-bearing Docker command loses host execution…

Deletes or force-pushes

Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository10 other files it gives its agents.
- nurec/libs/losses/AGENTS.mdAGENTS.md
- skills/nurec-onboard/CLAUDE.mdCLAUDE.md
- nurec-onboardSkill
- audit-datasetSkill
- build-ncore-converterSkill
- run-nurecSkill
- validate-ncore-dataSkill
- nurecSkill
- physical-ai-av-benchmarkSkill
- physical-ai-datasetsSkill

The file

We split this AGENTS.md into the rules it gives an agent

7 rules. About 0 are common, 1 is distinctive, and 6 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific
could be automatic: a setting could enforce it

Distinctive:
Always assign container names.
line 12
Very specific:
Run long agent-managed containers detached and inspect their state with separate, short-lived docker commands.
line 23
Very specific:
Never use system python or python3 to import NCore.
line 32
Very specific:
Do not use sudo for destructive commands.
line 37
Very specific:
Do not clear directories with variable-plus-glob forms such as rm -rf "$dir"/*.
line 38
Very specific:
Treat ownership recovery requiring sudo chown as an operator action.
line 41
Very specific:
Follow the selected SKILL.md as the operating contract.
line 46

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

<!--
SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
SPDX-License-Identifier: Apache-2.0
-->

AGENTS.md

Project-level conventions for autonomous workflow execution.

Docker

Always assign container names.
Preserve the resource, GPU, environment, user, and container-side mount flags in the workflow's Docker templates. Substitute only documented placeholders.
Issue each Docker command as a single docker invocation built from literal words. Docker runs on the host rather than in the command sandbox, and that applies only to a command in this shape. A chained, piped, redirected, wildcard-bearing, environment-assignment-prefixed, $(...)-bearing, or $VAR-bearing Docker command loses host execution and fails as a socket connection error rather than as a policy refusal, so resolve placeholders, uid, gid, and paths in separate commands and write the results in literally. -e VAR with no =value passes the value through from the environment.
Run long agent-managed containers detached and inspect their state with separate, short-lived docker commands. Commands presented for an operator to run must stay in the foreground.
Issue Docker commands separately from host filesystem setup, output reads, and cleanup. Docker access is outside the command sandbox's filesystem and network boundaries.

NCore

Never use system python or python3 to import NCore. Run NCore operations through the Bazel targets specified by the workflow.

Shell safety

Do not use sudo for destructive commands.
Do not clear directories with variable-plus-glob forms such as rm -rf "$dir"/*. Remove a validated, fully qualified directory and recreate it, or use find "$dir" -mindepth 1 -delete.
Treat ownership recovery requiring sudo chown as an operator action.

Workflow execution

Execute workflows autonomously without requesting input.
Follow the selected SKILL.md as the operating contract.
Execute workflow implementation inline. Do not delegate converter creation, validation, reconstruction, or diagnosis to subagents.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free