documentation-style

par nvidia

Style des documents publics NVCF : court, simple, Markdown ASCII sans gras, émojis, tiret cadratin ou tiret demi-cadratin. À utiliser lors de la modification de documents publics, READMEs, AGENTS.md, compétences d’agent,…

npx skills add https://github.com/nvidia/nvcf --skill documentation-style

Documentation and writing style

Purpose

Keep NVCF documentation plain, consistent, and useful for external readers. Write short docs an agent or contributor can act on without private context. Avoid typographic Unicode that breaks grep and diff tools, and avoid emphasis patterns that make docs noisy.

When to use

  • Any new or edited public Markdown in this repo.
  • AGENTS.md, CLAUDE.md, SKILL.md, READMEs, plans, and guides.
  • GitHub issue bodies, commit messages, and Pull Request descriptions.
  • Existing docs you are changing for another reason.

Instructions

  1. No markdown bold for emphasis. Do not use **...**, __...__, raw HTML <b>, or <strong>. Use headings, lists, and backticks for paths, commands, env vars, and identifiers.

  2. No emojis in documentation or user-facing prose.

  3. No em-dash (Unicode U+2014) or en-dash (U+2013). Split the thought into two sentences, or use a comma, semicolon, parenthesis, or colon.

  4. ASCII only in committed prose. Avoid smart quotes, ellipsis characters, non-breaking spaces, arrows, and other typographic Unicode. Prefer -> over an arrow glyph and ... over a single-character ellipsis. Code blocks and quoted fixtures are exempt.

  5. Be succinct. Prefer short sentences and direct wording. Cut filler and repetition. Keep one idea per sentence when possible.

  6. Prefer lists for short, scan-friendly structure. Use bullets for unordered items and numbered lists when sequence, priority, or stable rule references matter. Use tables only when columns make comparison, status, option tradeoffs, or matrix decisions easier. Do not turn a short list into a table.

  7. Be easy to understand. Define acronyms on first use if the audience is broad. Prefer concrete examples over abstract jargon.

  8. Write for GitHub readers. Use public names, public paths, and public behavior. Replace private evidence with the user-visible reason for the change.

  9. Keep agent guidance actionable. Say when to use a skill, what to inspect first, what command to run, and what done means. Do not add temporary branch status, Pull Request queues, merge-order notes, or private routing details.

  10. Avoid vague agent language. Remove filler such as "ensure", "leverage", "robust", "comprehensive", "where applicable", and "as needed" unless the term adds a concrete condition.

Examples

Inline bold label, replaced with a plain prefix:

Before: **Note:** restart after edit.
After:  Note: restart after edit.

Bold section header, replaced with a real heading:

Before: **Prerequisites**

        - Repository access to github.com/NVIDIA/nvcf.

After:  ## Prerequisites

        - Repository access to github.com/NVIDIA/nvcf.

Double hyphen used as a dash, replaced with a period:

Before: Services live upstream -- this repo mirrors them.
After:  Services live upstream. This repo mirrors them.

Long agent guidance, replaced with commands and durable criteria:

Before: Ensure agents comprehensively verify symlink fanout where applicable.

After:  Run `python3 ai-tooling/dev/hooks/validate-skill-fanout.py` after
        changing root skills. Fix every reported path before finishing.

Editing existing files

If a file already uses bold, emojis, em-dashes, en-dashes, or non-ASCII punctuation, normalize it when you touch that file for another reason, or when the user asks for a style pass. Do not expand scope into unrelated trees just to fix style.

Limitations

  • The rules target agent-edited prose in this repo. Upstream documentation keeps its own style until merged natively.
  • There is no automated linter for prose today. Review catches style issues.

Exceptions

  • Content you are quoting verbatim, such as upstream LICENSE text, cited errors, or user paste.
  • Generated files where the generator controls formatting. Fix style in the generator if needed.
  • Auto-generated changelogs or release notes assembled from external commit messages. Fix the source commits, not the rolled-up file.

Plus de skills de nvidia

compileiq-debug
nvidia
Utilisez quand quelque chose ne va pas : Search() bloque, toutes les évaluations retournent INVALID_SCORE, les scores ne s'améliorent pas, chaque configuration retourne le même nombre, erreurs ptxas…
create-github-pr
nvidia
Créer des pull requests GitHub en utilisant l'interface en ligne de commande gh. Utiliser lorsque l'utilisateur souhaite créer une nouvelle PR, soumettre du code pour révision, ou ouvrir une pull request. Mots-clés de déclenchement -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Analyse les autres problèmes ouverts pour trouver ceux qu’une PR donnée pourrait également corriger ou casser accidentellement. Génère des opportunités de correctifs adjacents et des risques de contradiction avec fichier:ligne…
fhir-basics
nvidia
Apprend aux agents comment fonctionnent les API FHIR R4, quelles ressources sont disponibles, comment les interroger avec des paramètres de recherche, et comment analyser correctement tous les formats de réponse…
compileiq-validate-result
nvidia
Utiliser APRÈS qu'une recherche soit terminée et AVANT de réclamer un accélérateur ou d'expédier un ACF. Charge le CSV dump_results, extrait les K meilleurs candidats (mono-objectif)…
changelog-audit
nvidia
Auditer le CHANGELOG.md de Warp avant une publication : récupérer les entrées perdues, trier par impact utilisateur, affiner le langage des entrées, ajuster les retours à la ligne et (en mode branche de publication) mettre à jour la comparaison…
maintain-dynamic-plugins
nvidia
Maintenir les chargeurs de plugins dynamiques NeMo Relay, les manifestes, les SDK natifs Rust, le protocole worker gRPC, le SDK worker Python, la documentation, les tests et la couverture du workflow de publication
dgx-diagnose
nvidia
Diagnostiquer les problèmes courants du DGX Station GB300 — plantages CUDA, ciblage incorrect du GPU, bugs de conteneur vLLM/SGLang, problèmes d'état MIG, erreurs NVLink/Fabric Manager,…