overthinkos/rocm
ov-layers/skills/rocm/SKILL.md
AMD ROCm runtime, OpenCL, and GPU compute support via system packages. Use when working with AMD GPU computing, ROCm, HIP, OpenCL, or AMD GPU passthrough in containers.
data-ai
Updated Apr 24, 2026
$ install --global
skillsauthnpx skillsauth add overthinkos/overthink-plugins rocmInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 24, 2026, 3:11 PM246.9s1 file scanned
SKILL.md
- name:
- rocm
- description:
- |
rocm -- AMD ROCm runtime and OpenCL
Layer Properties
| Property | Value |
|----------|-------|
| Install files | layer.yml |
Environment Variables
| Variable | Value |
|----------|-------|
| ROCM_PATH | /usr |
Runtime Environment (auto-detected)
| Variable | Source | Example |
|----------|--------|---------|
| HSA_OVERRIDE_GFX_VERSION | KFD topology sysfs | 10.3.0 (RDNA2), 11.0.0 (RDNA3) |
| DRINODE | DRM render node enumeration | /dev/dri/renderD128 (typical), /dev/dri/renderD129 (multi-GPU) |
Both variables are not baked into the layer — they are auto-detected from host state at runtime and injected as container environment variables via appendAutoDetectedEnv() in ov/devices.go. The same function is called by ov config, ov start, and ov shell, so interactive shells and deployed services see the identical env set.
HSA_OVERRIDE_GFX_VERSIONis read from/sys/class/kfd/kfd/topology/nodes/*/properties(gfx_target_version field).DRINODEis selected by walking/dev/dri/renderD*and picking the node that matches the AMD PCI device exposed to the container.
Override either with -e HSA_OVERRIDE_GFX_VERSION=X.Y.Z or -e DRINODE=/dev/dri/renderD129. See /ov:doctor (Hardware Detection) for how the probe runs on the host side, and /ov-layers:nvidia (DRINODE Auto-Injection) for the NVIDIA counterpart using the same mechanism.
Security
security:
group_add:
- keep-groups
Uses keep-groups to preserve host supplementary groups (video, render) inside the container. This is the standard approach across all layers -- Podman's keep-groups is mutually exclusive with explicit group names.
Packages
RPM (Fedora system repos): rocm-hip-runtime, rocm-opencl, rocm-clinfo, rocm-smi
Host Requirements
AMD GPU support requires:
/dev/kfddevice (auto-detected byov)/dev/dri/renderD*render nodes (auto-detected)- User in
videoandrendergroups (ov udev statusto check) amdgpukernel driver loaded
Run ov doctor to verify detection. Run ov udev install to set up device permissions.
Usage
# image.yml -- standalone AMD GPU image
my-amd-app:
base: fedora
layers:
- rocm
- my-app
Verification
# Check AMD GPU detected on host
ov doctor | grep "AMD GPU"
# Verify inside container
ov shell my-amd-app -c "clinfo --list"
ov shell my-amd-app -c "rocm-smi"
ov shell my-amd-app -c "echo \$HSA_OVERRIDE_GFX_VERSION"
Related Layers
/ov-layers:nvidia-- NVIDIA GPU counterpart (runtime libs + CDI), sharesappendAutoDetectedEnv()DRINODE injection/ov-layers:cuda-- NVIDIA CUDA toolkit (stacked on nvidia)/ov-layers:python-ml-- ML Python environment (currently depends on cuda; ROCm equivalent is a future direction)
Related Commands
/ov:doctor-- Host AMD GPU detection (/dev/kfd, render nodes, driver status)/ov:shell-- Interactive shells receive the same auto-detected HSA_OVERRIDE_GFX_VERSION + DRINODE envs/ov:udev-- Device permission management for/dev/kfdand/dev/dri/renderD*/ov:config-- Runtime GPU env injection at deployment time (same auto-detect path)/ov:start-- Runtime GPU env injection at service start time
Used In Images
Not directly used in any current image definition. Available as a standalone layer for AMD GPU support. The NVIDIA base image (/ov-images:nvidia) is the currently-shipped GPU image; an AMD counterpart can be composed by substituting this layer.
When to Use This Skill
Use when the user asks about:
- AMD GPU support in containers
- ROCm, HIP, or OpenCL setup
/dev/kfddevice accessHSA_OVERRIDE_GFX_VERSIONconfiguration- AMD GPU passthrough or compute
- The rocm layer or AMD GPU computing
- Comparing AMD vs NVIDIA GPU support
Related
/ov:layer— layer authoring reference (layer.ymlschema, task verbs, service declarations)/ov:test— declarative testing (tests:block,ov image test,ov test)
Related Skills
overthinkos/plugin
tools
VerifiedTrustedCommunity
Use when authoring or modifying a charly PLUGIN — a candy with a `plugin:` block that contributes Providers (verbs/kinds/deploy-targets/steps/builders/commands), its own CUE schema, builtin (compiled-in) or external (out-of-tree git repo). Covers the unified Provider model, the per-plugin CUE-schema contract (single source → Go params for dev + schema-over-Describe RPC for runtime), the SDK, and the loader.
SKILL.mdUpdated Jun 22, 2026
overthinkos/cue
tools
VerifiedTrustedCommunity
The CUE data-validation / configuration CLI (cue), pinned to v0.16.1. Use when working with the cue candy, installing the cue binary into a box or onto a target:local dev host, or running the offline schema-vendoring pipeline that feeds charly's egress validation.
SKILL.mdUpdated Jun 16, 2026
overthinkos/egress
tools
VerifiedTrustedCommunity
CUE EGRESS validation — validating (and, where it adds value, generating) the config files charly WRITES to a system BEFORE the bytes hit disk. MUST be invoked before working on charly/egress.go, the vendored schemas under candy/plugin-egress/egress-schemas/vendor/, the ValidateEgress / registerVendoredEgressKind path, the offline `task cue:vendor` pipeline, or adding an egress schema for any written artifact (cloud-init, k8s manifests, traefik routes, runtime config, install ledger, systemd/quadlet units, ssh_config, libvirt XML).
SKILL.mdUpdated Jun 16, 2026
overthinkos/check-k8s
tools
VerifiedTrustedCommunity
Kubernetes cluster-probe declarative check verb — the `kube:` check verb (nodes, pods, ingress, storage class, addon health, apply/delete, and arbitrary resource GETs) served out-of-process by the candy/plugin-kube plugin (vendored client-go; no external kubectl required).
SKILL.mdUpdated Jun 14, 2026