local-llm-expert

Master local LLM inference, model selection, VRAM optimization, and local deployment using Ollama, llama.cpp, vLLM, and LM Studio. Expert in quantization formats (GGUF, EXL2) and local AI privacy.

Quality

54%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/local-llm-expert/SKILL.md

Quality

Content

31%Scale 1-5

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

This skill reads more like a persona/role-play prompt than an actionable skill document. It extensively describes what Claude should know and how it should behave, but provides almost no concrete, executable guidance—no actual commands, code snippets, configuration files, or specific examples. The content is verbose with capability descriptions that Claude doesn't need to be told about in this manner.

Suggestions

Replace the 'Example Interactions' parenthetical descriptions with actual complete responses including real CLI commands (e.g., `ollama pull llama3:8b-instruct-q4_K_M && ollama run llama3:8b-instruct-q4_K_M`), real Modelfile content, and real Python code.

Add a concrete VRAM calculation table or formula with worked examples instead of just describing the formula in prose (e.g., 'Llama 3 8B at Q4_K_M: 8B × 4.5bpw / 8 ≈ 4.5GB + ~1GB KV cache = ~5.5GB').

Remove persona-style sections ('Behavioral Traits', 'Knowledge Base') and convert their useful content into actionable reference material—e.g., a chat template cheat sheet with exact token formats for each model family.

Create bundle files for detailed reference content (quantization format comparison table, chat template examples, engine-specific configuration guides) and link to them from a streamlined SKILL.md overview.

Dimension	Reasoning	Score
Conciseness	Heavily padded with sections that describe Claude's persona and capabilities rather than providing actionable instructions. Sections like 'Behavioral Traits', 'Knowledge Base', 'Capabilities' read as role-play descriptions rather than efficient skill content. Extensive over-explanation of concepts Claude already knows (what VRAM is, what quantization is, what these tools do).	2 / 5
Actionability	Despite covering a broad domain, the skill contains zero executable code, no actual CLI commands, no concrete configuration examples, and no copy-paste ready snippets. The 'Example Interactions' section describes what responses would contain in parenthetical notes rather than actually providing them. Everything remains at the level of vague direction ('provides exact CLI command') without ever giving one.	2 / 5
Workflow Clarity	The 'Instructions' and 'Response Approach' sections provide a reasonable high-level sequence (assess hardware → select model → provide commands → format template → optimize), but there are no validation checkpoints, no error recovery steps, and no concrete verification steps. The OOM troubleshooting example hints at a feedback loop but doesn't formalize it.	3 / 5
Progressive Disclosure	The content is a monolithic file with no references to supporting files, no links to external resources, and no bundle files. All content is inlined in a single document with many sections that could benefit from being split out (e.g., quantization format reference, chat template examples, VRAM calculation tables). The flat structure makes navigation difficult.	2 / 5
	Total	9 / 20 Passed

Description

77%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 excels at naming specific tools, formats, and technical domains that create a clear niche with excellent trigger terms. However, it uses 'Expert in' phrasing rather than action-oriented language, and critically lacks a 'Use when...' clause that would help Claude know when to select this skill. The capabilities are listed more as topics of expertise than concrete actions the skill performs.

Suggestions

Add an explicit 'Use when...' clause, e.g., 'Use when the user asks about running models locally, setting up Ollama, optimizing VRAM usage, choosing quantization formats, or troubleshooting local AI inference.'

Reframe from expertise-listing to action-oriented language: instead of 'Master local LLM inference' and 'Expert in quantization formats', use verbs like 'Guides setup of local LLM inference', 'Recommends quantization formats', 'Troubleshoots VRAM issues'.

Dimension	Reasoning	Score
Specificity	Lists several specific capabilities: local LLM inference, model selection, VRAM optimization, local deployment, quantization formats. However, it reads more like a list of topics/domains than concrete actions (e.g., 'configure', 'troubleshoot', 'benchmark' would be more action-oriented).	4 / 5
Completeness	Has a clear 'what' (local LLM inference, model selection, VRAM optimization, deployment with specific tools) but completely lacks a 'when' clause. There is no explicit 'Use when...' guidance for Claude to know when to select this skill.	3 / 5
Trigger Term Quality	Excellent coverage of natural keywords users would say: Ollama, llama.cpp, vLLM, LM Studio, GGUF, EXL2, VRAM, quantization, local LLM, local AI, local deployment. These are precisely the terms users searching for local AI help would use.	5 / 5
Distinctiveness Conflict Risk	Highly distinctive niche focused specifically on local LLM deployment with named tools (Ollama, llama.cpp, vLLM, LM Studio) and specific formats (GGUF, EXL2). Very unlikely to conflict with other skills.	5 / 5
	Total	17 / 20 Passed

Validation

90%

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 — 10 / 11 Passed

Validation for skill structure

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

	Total	10 / 11 Passed

Repository: administrakt0r/AI-Agents-Safe-Coding-Skills
Path: skills/local-llm-expert/SKILL.md
Commit: 57c135b

Reviewed: 4 days 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.