Skill audit report
serving-llms-on-epyc Audit report.
Serves an LLM on a supported AMD EPYC server CPU using vLLM with zentorch, in Docker, Podman, or conda. Use for "vLLM on CPU", "zentorch serving", or an EPYC CPU endpoint, including on a host that also has AMD Instinct GPUs. Detects the EPYC generation, validates the runtime, checks model support and RAM fit, sizes threads/KV/NUMA, confirms the plan, launches, and verifies the endpoint. Runs one instance on one socket and its memory. Reports and stops on failure; does not retry or debug. Use serving-llms-on-instinct when the endpoint should run on a GPU. Excludes multi-node, EPYC 4000, and pre-Zen4 EPYC without AVX-512.
OpenAgentSkill Trust Score
OpenAgentSkill Trust Score
The Trust Score helps an agent decide whether a skill is safe enough to shortlist before installation.
GitHub adoption
INFO62
332 GitHub stars
Stars/forks activity
WARN57
332 stars, 30 forks; issue activity unavailable in current metadata
Recent maintenance
PASS88
2mo since push
License clarity
PASS86
MIT
README/SKILL.md completeness
PASS86
Metadata includes enough usage and workflow context
Dependency/runtime risk
WARN44
command execution surface, credential or environment access
Install availability
PASS92
npx skills add amd/skills --skill serving-llms-on-epyc
Install command safety
PASS92
standard package or runtime install path
Permission surface
FAIL22
secrets or environment access, shell or command execution
Repository evidence
PASS86
https://github.com/amd/skills/tree/main/skills/serving-llms-on-epyc
Review status
INFO66
AI review data available
Agent Proven outcomes
INFO54
No agent outcome data yet
Checks
Install and adoption review
Install path
92
npx skills add amd/skills --skill serving-llms-on-epyc
Repository
88
https://github.com/amd/skills/tree/main/skills/serving-llms-on-epyc
License
86
MIT
Maintenance
88
2mo since push
AI review
55
The skill relies on external resources (Docker Hub image, Hugging Face model registry) which could introduce supply-chain risks if the image or registry is compromised; however, the skill pins a specific image and version, mitigating this.
README/SKILL.md completeness
86
Usable description available
Dependency risk
44
command execution surface, credential or environment access
Install command safety
92
standard package or runtime install path
Permission surface
22
secrets or environment access, shell or command execution
Stars/forks activity
57
332 stars, 30 forks; issue activity unavailable in current metadata
Adoption
68
332 GitHub stars
Financial decision safety
58
Research-only use: do not treat output as financial advice or execute a position without human approval.
Warnings
- Dependency or permission surface needs review
- Permission surface may require sandboxing
- Financial research output is not financial advice; require human review before any live investment decision
- The skill relies on external resources (Docker Hub image, Hugging Face model registry) which could introduce supply-chain risks if the image or registry is compromised; however, the skill pins a specific image and version, mitigating this.
- The skill assumes the agent has permission to run docker/podman commands without sudo; this is a stated prerequisite, but the skill does not verify user permissions beyond runtime detection.
- Financial research output is not financial advice; require human review before any live investment decision.
- Quality score needs review
- Permission surface needs review: secrets or environment access, shell or command execution
- Stars/forks activity: 332 stars, 30 forks; issue activity unavailable in current metadata
- Dependency/runtime risk: command execution surface, credential or environment access
- Permission surface: secrets or environment access, shell or command execution
Method
This report combines public metadata, AI review output, repository freshness, install readiness, OpenAgentSkill events, quality scoring, trust checks, and the agent safety gate. It is not a full source-code security review.