nvcf / ai-tooling
AGENTS.md261 starsChanged 9 days agoApache-2.0View on GitHub
This subtree contains public Agent Skills for NVIDIA Cloud Functions (NVCF). All skills must comply with the Agent Skills specification plus the repo-specific public skill metadata contract. This repository is compatible with the Vercel Skills CLI. Use this file as the authoring contract for every public skill in user/skills/ and dev/skills/. Keep private skills in the private subtree. When adding or editing a skill, keep the external skill spec fields at the top level of the frontmatter, retain the required…
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 31,700 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 repository133 other files it gives its agents.
- AGENTS.mdAGENTS.md
- deploy/helm/cassandra/AGENTS.mdAGENTS.md
- deploy/helm/cert-manager/AGENTS.mdAGENTS.md
- deploy/helm/cloud-functions/AGENTS.mdAGENTS.md
- deploy/helm/cloud-tasks/AGENTS.mdAGENTS.md
- deploy/helm/encrypted-secret-store/AGENTS.mdAGENTS.md
- deploy/helm/icms/AGENTS.mdAGENTS.md
- deploy/helm/llm-request-router/AGENTS.mdAGENTS.md
- deploy/helm/nats/AGENTS.mdAGENTS.md
- deploy/helm/notary/AGENTS.mdAGENTS.md
- deploy/helm/nvcf-pki/AGENTS.mdAGENTS.md
- deploy/helm/openbao/AGENTS.mdAGENTS.md
- deploy/helm/vanity-gateway/AGENTS.mdAGENTS.md
- deploy/releases/AGENTS.mdAGENTS.md
- deploy/stacks/AGENTS.mdAGENTS.md
- deploy/stacks/observability/AGENTS.mdAGENTS.md
- deploy/stacks/stargate-dev/AGENTS.mdAGENTS.md
- deploy/stacks/stargate-dev/benchmark/AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- docs/overview/samples/scripts/AGENTS.mdAGENTS.md
- examples/AGENTS.mdAGENTS.md
- infra/cassandra/AGENTS.mdAGENTS.md
- infra/openbao/AGENTS.mdAGENTS.md
- infra/openbao/plugins/vault-plugin-secrets-jwt/AGENTS.mdAGENTS.md
- infra/openbao/tests/AGENTS.mdAGENTS.md
- migrations/cassandra/AGENTS.mdAGENTS.md
- migrations/openbao/AGENTS.mdAGENTS.md
- src/compute-plane-services/byoo-otel-collector/AGENTS.mdAGENTS.md
- src/compute-plane-services/ess-agent/AGENTS.mdAGENTS.md
- src/compute-plane-services/image-credential-helper/AGENTS.mdAGENTS.md
- src/compute-plane-services/nvca/AGENTS.mdAGENTS.md
- src/compute-plane-services/nvsnap/docs/AGENTS.mdAGENTS.md
- src/compute-plane-services/nvstor/AGENTS.mdAGENTS.md
- src/compute-plane-services/request-trace-uploader/AGENTS.mdAGENTS.md
- src/control-plane-services/admin-token-issuer-proxy/AGENTS.mdAGENTS.md
- src/control-plane-services/api-keys/AGENTS.mdAGENTS.md
- src/control-plane-services/cloud-functions/AGENTS.mdAGENTS.md
- src/control-plane-services/cloud-tasks/AGENTS.mdAGENTS.md
- src/control-plane-services/encrypted-secret-store/AGENTS.mdAGENTS.md
- src/control-plane-services/function-autoscaler/AGENTS.mdAGENTS.md
- src/control-plane-services/function-autoscaler/local_env/AGENTS.mdAGENTS.md
- src/control-plane-services/instance-cluster-management/AGENTS.mdAGENTS.md
- src/control-plane-services/nats-auth-callout/AGENTS.mdAGENTS.md
- src/control-plane-services/notary/AGENTS.mdAGENTS.md
- src/invocation-plane-services/grpc-proxy/AGENTS.mdAGENTS.md
- src/invocation-plane-services/http-invocation/AGENTS.mdAGENTS.md
- src/invocation-plane-services/llm-api-gateway/AGENTS.mdAGENTS.md
- src/invocation-plane-services/ratelimiter/AGENTS.mdAGENTS.md
- src/invocation-plane-services/vanity-gateway/AGENTS.mdAGENTS.md
- src/libraries/go/lib/AGENTS.mdAGENTS.md
- src/libraries/go/worker/AGENTS.mdAGENTS.md
- src/libraries/java/nv-boot-parent/AGENTS.mdAGENTS.md
- src/libraries/python/nv-cloud-function-helpers/AGENTS.mdAGENTS.md
- src/libraries/rust/nvcf-info/AGENTS.mdAGENTS.md
- src/libraries/rust/stargate/AGENTS.mdAGENTS.md
- src/uis/nvcf-ui/AGENTS.mdAGENTS.md
- src/uis/nvcf-ui/helm/AGENTS.mdAGENTS.md
- tests/bdd/AGENTS.mdAGENTS.md
- tools/AGENTS.mdAGENTS.md
- tools/docs-version-sync/AGENTS.mdAGENTS.md
- The first 60 shown.
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 0 are common, 7 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 7
Use this file as the authoring contract for every public skill in user/skills/ and dev/skills/.
- Very specific: line 9
When adding or editing a skill, keep the external skill spec fields at the top level of the frontmatter, retain the required repo-specific …
- Distinctive: line 108
Must not start or end with a hyphen
- Very specific: line 109
Must not contain consecutive hyphens (--)
- Distinctive: line 110
Must match the parent directory name exactly
- Distinctive: line 117
When to use it (trigger phrases, keywords)
- Very specific: line 127
Keep main SKILL.md under 500 lines
- Distinctive: line 132
Document common edge cases
- Distinctive: line 136
Before You Start - list required config, binaries, and access.
- Distinctive: line 143
Use relative paths from the skill root:
- Distinctive: line 153
Before committing changes, verify:
≈ “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.
Files
Agent Skills Repository Guidelines
This subtree contains public Agent Skills for NVIDIA Cloud Functions (NVCF). All skills must comply with the Agent Skills specification plus the repo-specific public skill metadata contract. This repository is compatible with the Vercel Skills CLI.
Instructions
Use this file as the authoring contract for every public skill in user/skills/ and dev/skills/. Keep private skills in the private subtree.
When adding or editing a skill, keep the external skill spec fields at the top level of the frontmatter, retain the required repo-specific fields under metadata, and include a concise ## Instructions section in the body.
Skill Structure
Source skills live under user/skills/ or dev/skills/. Each skill must be in its own subdirectory containing a SKILL.md file. The directory name must match the name field in the frontmatter.
user/skills/
- user-skill-name/
- SKILL.md
dev/skills/
- dev-skill-name/
- SKILL.mdPublic hook source scripts live under dev/hooks/. Do not place hook implementation scripts directly under .cursor/hooks/, .codex/hooks/, or .claude/hooks/; those root directories are fanouts only and must contain matching symlinks to dev/hooks/.
SKILL.md Format
Complete Frontmatter Template
---
name: skill-name
description: Clear description of what the skill does and when to use it. Include specific keywords to help agents identify relevant tasks.
license: MIT
compatibility: Requires NGC CLI installed and configured
author: "nvcf-core-eng <nvcf-core-eng@exchange.nvidia.com>"
version: "1.0.0"
tags:
- ngc
- nvcf
- cli
- cloud-infrastructure
- skill
tools:
- Shell
- Read
- Write
metadata:
internal: false
author: "nvcf-core-eng <nvcf-core-eng@exchange.nvidia.com>"
version: "1.0"
tags:
- ngc
- nvcf
- relevant-tag
languages:
- bash
frameworks:
- ngc-cli
domain: cloud-infrastructure
---Required Fields
| Field | Type | Constraints |
name | string | 1-64 chars, lowercase, alphanumeric + hyphens, no leading/trailing/consecutive hyphens, must match directory name |
description | string | 1-1024 chars, must describe what AND when to use, include keywords for discoverability |
author | string | Author name or team |
version | string | Semantic version string |
tags | array | Categorization tags (recommend 1-5 items for top-level spec compatibility) |
tools | array | Agent tools required by the skill (for example Shell, Read, Write) |
Optional Fields
| Field | Type | Description |
license | string | License name or reference to bundled license file |
compatibility | string | Max 500 chars. Environment requirements |
metadata | object | Repo-specific public skill metadata fields (see below) |
Public Skill Metadata Fields (Required)
These fields must be placed inside the metadata object:
| Field | Type | Description | Example |
internal | boolean | Must be false for public skills in this subtree | false |
author | string | Author name and NVIDIA email | "Your Name <your-email@nvidia.com>" |
version | string | Skill version | "1.0" |
tags | array | List of searchable tags | ["ngc", "nvcf", "cli"] |
languages | array | Programming languages this skill applies to | ["bash", "python"] |
frameworks | array | Frameworks or libraries this skill covers | ["ngc-cli"] |
domain | string | Domain or area of expertise | "cloud-infrastructure" |
Fields NOT to Include
alwaysApply- This is for rules, not skillsglobs- This is for rules, not skills
Skills are invoked on-demand, not automatically applied like rules.
Name Validation Rules
- Lowercase letters, numbers, and hyphens only (
a-z,0-9,-) - Must not start or end with a hyphen
- Must not contain consecutive hyphens (
--) - Must match the parent directory name exactly
Description Best Practices
Include:
- What the skill does (actions it enables)
- When to use it (trigger phrases, keywords)
Example:
description: Manage NVCF clusters via NGC CLI. Register, list, and delete clusters for function and task deployments. Use when registering clusters, managing cluster configurations, or when the user mentions ngc cf cluster, NVCF clusters, or cluster registration.Body Content Guidelines
- Keep main
SKILL.mdunder 500 lines - Move detailed reference material to
references/directory - Include a concise
## Instructionssection near the top of the body - Include step-by-step instructions
- Provide command examples
- Document common edge cases
Standard Sections for CLI Skills
- Before You Start - list required config, binaries, and access.
- Core Commands - primary operations with examples.
- Examples - real-world usage scenarios.
- Additional Resources - links to detailed docs or reference files.
File References
Use relative paths from the skill root:
For detailed examples, see [examples.md](examples.md).
For reference material, see [references/REFERENCE.md](references/REFERENCE.md).
Run the helper script: `python scripts/helper.py`Validation Checklist
Before committing changes, verify:
namefield matches directory namenameis lowercase with hyphens onlydescriptionincludes what AND when to usecompatibilityfield is present- Top-level
authoris present - Top-level
versionis present and semantic - Top-level
tagsis present - Top-level
toolsis present metadata.authorincludes name and NVIDIA emailmetadata.versionis presentmetadata.tagsarray is populatedmetadata.languagesarray is populatedmetadata.frameworksarray is populatedmetadata.domainis set- File is under 500 lines
- All relative links are valid
- No
alwaysApplyorglobsfields present - If editing
nvcf-self-managed-cliornvcf-self-managed-installationsource files: rungo generate ./internal/agentskill/...insrc/clis/nvcf-cliand stageskilldata_generated.go
Skill Inventory
Root AGENTS.md is the fanout index for public and private dev skills exposed
through .cursor/skills/, .codex/skills/, and .claude/skills/. Keep this
file focused on the public skill authoring contract. Use README.md for the
public catalog users browse in this subtree.
References
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
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.

