Tessl • The management layer for AI agent skills

SkillPublic

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

62%

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 ./plugins/AI-Agents-Safe-Coding-Skills/skills/local-llm-expert/SKILL.md

The canonical home for this skill is local-llm-expert in administrakt0r/AI-Agents-Safe-Coding-Skills

Quality

Content

42%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.

A well-organized persona skill with a coherent workflow and some genuinely concrete details (VRAM formula, flags, template strings), but it is structurally redundant, promises rather than provides executable commands, and lacks validation checkpoints. Consolidating duplicated sections and including real command/code examples are the highest-leverage fixes.

Suggestions

Collapse the duplicated sections — merge "## Purpose" into the opening paragraph, fold "## Knowledge Base" into "## Capabilities", and merge "## Response Approach" with "## Instructions" — to cut roughly a third of the body.

Include actual executable artifacts instead of parenthetical descriptions: a full `ollama run` / Modelfile example, a vLLM server launch command, and a short Python client snippet in the Example Interactions.

Add validation checkpoints to the workflow, e.g. "verify the calculated VRAM (model + KV cache) fits the user's card before recommending" and "confirm the chat template renders correctly before handing off commands".

Dimension	Reasoning	Score
Conciseness	Noticeably verbose with several redundant sections: the opening persona paragraph and "## Purpose" repeat the same claims nearly verbatim, "## Knowledge Base" recapitulates "## Capabilities" (GGUF catalog, Ollama API, benchmarks), and "## Response Approach" duplicates the "## Instructions" sequence. Not 3 because the redundancy is structural and pervasive rather than a few trimmable sentences.	2 / 5
Actionability	Some concrete guidance exists (the VRAM formula "Parameters * Bits-per-weight / 8", flags like `-ngl` and `num_ctx`, the exact ChatML string), but no complete executable commands or code blocks appear anywhere — the "Example Interactions" describe outputs in parentheses ("provides `ollama run` command") instead of giving them. Not 4 because key executable details are consistently missing.	3 / 5
Workflow Clarity	A clear 5-step sequence is present (confirm hardware → recommend model/quant → provide commands → supply template → emphasize privacy), but there are no validation checkpoints such as verifying the calculated VRAM fits or confirming the model launches. Not 4 because checkpoints are absent entirely, not merely implicit; not 2 because the sequence itself is coherent and well-ordered.	3 / 5
Progressive Disclosure	The single file is well-sectioned, but at ~75 lines (over the 50-line simple-skill threshold) it inlines material that belongs in reference files: the claimed "complete catalog of GGUF formats and their bitrates", chat-template formats, and benchmark tables are promised in the body rather than delivered or split out. Not 4 because the organization gaps are more than minor given the duplication across sections.	3 / 5
	Total	11 / 20 Passed

Description

66%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.

A specific, well-keyworded description with a clear niche, undermined by the complete absence of an explicit "Use when..." trigger clause. Adding trigger guidance and a few natural user phrasings would lift it substantially.

Suggestions

Add an explicit trigger clause, e.g. "Use when the user wants to run LLMs locally, offline, or on consumer hardware, or mentions Ollama, llama.cpp, vLLM, GGUF, or VRAM limits."

Include natural user phrasings such as "run models locally", "offline AI", and "open-weight models" to broaden trigger-term coverage.

Replace role-claim framing ("Master...", "Expert in...") with concrete actions like "Calculates VRAM requirements, recommends quantization formats, and writes launch commands".

Dimension	Reasoning	Score
Specificity	The description lists several concrete action areas ("model selection, VRAM optimization, and local deployment") and names specific tools (Ollama, llama.cpp, vLLM, LM Studio). It falls short of the 5 anchor because "Master..." and "Expert in..." are role-claim framing rather than a comprehensive list of concrete actions.	4 / 5
Completeness	The "what" is clear and multi-part, but there is no "Use when..." clause or equivalent trigger guidance, which caps completeness at 3 per the rubric guidelines. Not 2 because the "what" is concrete and specific rather than vague.	3 / 5
Trigger Term Quality	Good natural keyword coverage: "local LLM", "VRAM", "Ollama", "llama.cpp", "vLLM", "GGUF", "EXL2", "local AI privacy" — terms users would actually say. Not 5 because common phrasings like "run models locally", "offline AI", or "open-weight models" are missing.	4 / 5
Distinctiveness Conflict Risk	The local-inference niche (local deployment, GGUF/EXL2 quantization, named local engines) is clearly distinct from cloud-API skills. Minor overlap risk remains with cloud LLM or fine-tuning skills, so not 5.	4 / 5
	Total	15 / 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: administrakt0r/AI-Agents-Safe-Coding-Skills
Path: plugins/AI-Agents-Safe-Coding-Skills/skills/local-llm-expert/SKILL.md
Revision: 64adc55

Reviewed: 14 days ago

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.

local-llm-expert

Table of Contents

Is this your skill?