llama-cpp

Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.

Quality

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

80%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable and well-structured with progressive disclosure through three real reference files. Its main weakness is workflow clarity: batch and large-model operations lack validation checkpoints or feedback loops.

Suggestions

Add explicit validation/verification steps to the batch processing and large-model deployment workflows (e.g., verify output count matches input, or check perplexity before deploying a quantized model).

Add a brief error-recovery feedback loop for the server-mode and quantization examples (e.g., if OOM, reduce -ngl or context) to lift workflow clarity above the batch cap.

Trim the inline CPU/GPU benchmark tables since equivalent detail lives in optimization.md, keeping the overview leaner.

Dimension	Reasoning	Score
Conciseness	The body is command-driven and assumes competence with minimal concept explanation, but inline benchmark tables and some hardware sections pad beyond what is strictly needed, fitting the 'efficient, minor over-explanation' anchor.	4 / 5
Actionability	Copy-paste-ready bash commands cover installation, download, inference, server mode, and quantization selection, with the OpenAI-compatible curl example matching the fully-executable anchor.	5 / 5
Workflow Clarity	Quick start sections are sequenced but the batch processing and large-model sections perform batch/large operations with no validation or verify step, capping workflow clarity at 3 per the batch-operation guideline.	3 / 5
Progressive Disclosure	SKILL.md is a concise overview with three well-signaled, one-level-deep references (quantization.md, server.md, optimization.md) that all exist as real files, matching the clear-overview-with-easy-navigation anchor.	5 / 5
	Total	17 / 20 Passed

Description

83%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is concrete and complete, giving both capabilities and an explicit 'Use for' trigger focused on the non-NVIDIA niche. It is well-distinguished from sibling inference skills, with only minor keyword coverage gaps.

Dimension	Reasoning	Score
Specificity	Names the domain and several concrete capabilities ('Runs LLM inference', 'Supports GGUF quantization (1.5-8 bit)', 'reduced memory', '4-10x speedup') with only minor coverage gaps, fitting the 'lists several specific actions' anchor.	4 / 5
Completeness	Clearly states what ('Runs LLM inference on CPU, Apple Silicon, and consumer GPUs', GGUF quantization) and explicitly answers when via 'Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable', matching the explicit what-and-when anchor.	5 / 5
Trigger Term Quality	Good natural keyword coverage ('edge deployment', 'M1/M2/M3 Macs', 'AMD/Intel GPUs', 'when CUDA is unavailable') that a user might say, though a few common synonyms are missing, matching the 'good coverage, a few terms missing' anchor.	4 / 5
Distinctiveness Conflict Risk	The 'without NVIDIA hardware' / non-CUDA niche is distinct from vLLM/TensorRT-LLM territory with minimal overlap risk, fitting the 'mostly distinct, minor overlap' anchor rather than the fully distinctive 5.	4 / 5
	Total	17 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: OpenLAIR/dr-claw
Path: skills/inference-serving/llama-cpp/SKILL.md
Commit: 4b35756

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.