agentleFS
NVIDIAKnown publisherNVIDIA/nvcf/src/compute-plane-services/nvsnap/docs/AGENTS.md

nvcf / nvsnap

AGENTS.md261 starsChanged 9 days agoApache-2.0View on GitHub

Rules for anyone, person or agent, who adds to or changes nvsnap's documentation under this directory. make check-docs (and the nvsnap docs workflow) enforces the ones a script can check; the rest are for review. - Where a document goes - What every document does - Numbers - Figures - Generated pages - Before you push <!-- /toc --> The top level holds only START-HERE, ARCHITECTURE, RESULTS, GLOSSARY, REFERENCE, README and this file. A new document goes into README.md in…

The file

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

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When it finds another document that disagrees with the code, adds a row to ARCHITECTURE.md §14 or fixes that document in the same PR.

    line 52
  2. Distinctive:

    Every number names the hardware, the date and the nvsnap build ("not recorded" when the record does not say), and links the record it comes …

    line 58

≈ “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 - nvsnap documentation

Rules for anyone, person or agent, who adds to or changes nvsnap's documentation under this directory. make check-docs (and the nvsnap docs workflow) enforces the ones a script can check; the rest are for review.

<!-- toc: generated by scripts/docs/docs-toc.py; edit the headings, not this list -->

Contents

  • Where a document goes
  • What every document does
  • Numbers
  • Figures
  • Generated pages
  • Before you push
<!-- /toc -->

Where a document goes

KindWhere
Entry pointsSTART-HERE.md, then README.md (the index), then GLOSSARY.md
How nvsnap worksARCHITECTURE.md: the one technical document. A behaviour change updates its section in the same PR
Every numberRESULTS.md, with the record it comes from under results/
Working notes behind an architecture sectiondesign/, each opening with a status line
Recipes: install, operate, developoperate/
A dated run: a benchmark, a test campaignresults/; a record, not maintained
A superseded documentNowhere: delete it (git history keeps it), after moving any still-true rule into ARCHITECTURE.md and any result that has no other record into RESULTS.md

The top level holds only START-HERE, ARCHITECTURE, RESULTS, GLOSSARY, REFERENCE, README and this file. A new document goes into README.md in the same commit that adds it, with what it answers and its status.

What every document does

  • Uses the glossary's words with their meanings, and none of its retired words. A new term goes into the glossary in the same commit.
  • Names the code behind a claim by path and symbol (internal/agent/criu_group.go captureOptedIn), not by line number.
  • States a rule as the condition and the code that enforces it; states a limit as what is not done.
  • Describes a cluster by its hardware ("4x GB300, driver 610"; "H100, EKS with NVMesh storage"), never by its name, and names no internal host, address or personal path.
  • Cites no file that is not in this repository, such as a personal agent instructions file.
  • When it finds another document that disagrees with the code, adds a row to ARCHITECTURE.md §14 or fixes that document in the same PR.

Numbers

Every number names the hardware, the date and the nvsnap build ("not recorded" when the record does not say), and links the record it comes from. A number with no record is left out. A number found wrong is withdrawn in RESULTS.md, in place, with the reason; it is not silently replaced.

Figures

Review-grade figures follow the NVIDIA Cloud Architecture Diagrams style, through the nvaf-architecture-diagrams skill in .agents/skills/ (copied unchanged from the nvaf repository, as flashbox and bedrock do). Each figure is an SVG drawn by its own script under figures/scripts/, with a light and a dark PNG rendered beside it by figures/scripts/render.sh; a document embeds the SVG with ![alt](figures/<name>.svg). Look at both PNGs before committing, and commit the script, the SVG and the PNGs together. Mermaid is for quick inline sketches. How to regenerate: figures/README.md.

Generated pages

  • REFERENCE.md is generated from the Go sources by scripts/docs/reference.py; rerun it after changing an annotation, an environment variable or an agent route.
  • Tables of contents are generated by scripts/docs/docs-toc.py between their marker comments; edit the headings, never the list.

Before you push

make -C src/compute-plane-services/nvsnap check-docs

It fails on a broken link, heading anchor or doc path (in documents, code, scripts and the chart), a stale table of contents, a REFERENCE.md that no longer matches the code, and a figure that no longer matches its script. The repository's markdown-lint workflow checks style.

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.