model-quantization

Expert skill for AI model quantization and optimization. Covers 4-bit/8-bit quantization, GGUF conversion, memory optimization, and quality-performance tradeoffs for deploying LLMs in resource-constrained JARVIS environments.

Quality

72%

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Fix and improve this skill with Tessl

tessl review fix ./skills/model-quantization/SKILL.md

Quality

Content

77%

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

This is a comprehensive, highly actionable skill with excellent executable code examples and clear TDD workflow with validation checkpoints. The main weaknesses are verbosity (redundant sections, repeated concepts, inconsistent numbering) and the monolithic structure that doesn't follow through on the promised split-file organization. The security-conscious approach with checksum verification is a strength.

Suggestions

Remove redundant sections and fix inconsistent numbering (sections jump from 5.1/3.2 to 4, 5, 8, 13) - consolidate into a cleaner structure

Actually split detailed implementation patterns into the referenced 'references/' directory and keep SKILL.md as a concise overview with links

Remove explanatory text about what quantization is and what different levels mean conceptually - Claude knows this; keep only the decision-relevant table

Dimension	Reasoning	Score
Conciseness	The skill contains some unnecessary verbosity, including redundant sections (e.g., section numbering restarts at 3.2, 4, 5), repeated concepts across overview and summary, and explanatory text that Claude would already know (e.g., explaining what quantization levels mean conceptually). However, the code examples are generally lean.	2 / 3
Actionability	Provides fully executable Python code with complete implementations for quantization, benchmarking, model selection, and conversion. Code is copy-paste ready with proper imports, error handling, and realistic usage patterns.	3 / 3
Workflow Clarity	Clear TDD workflow with explicit steps (write test → implement → refactor → verify). Includes validation checkpoints throughout (checksum verification, integrity checks before loading, benchmark thresholds). The pre-deployment checklist provides explicit verification gates.	3 / 3
Progressive Disclosure	The header mentions 'See references/ for detailed implementations' but the skill itself is monolithic with extensive inline code that could be split. The content is reasonably organized with sections, but there's no actual linking to referenced files and the document is quite long with all patterns inline.	2 / 3
	Total	10 / 12 Passed

Description

67%

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 demonstrates strong technical specificity and carves out a clear niche for model quantization tasks. However, it lacks an explicit 'Use when...' clause and could benefit from more natural trigger terms that users might actually say when needing this skill.

Suggestions

Add an explicit 'Use when...' clause, e.g., 'Use when the user asks about compressing models, reducing VRAM usage, converting to GGUF, or running LLMs locally.'

Include more natural user-facing trigger terms like 'compress model', 'reduce model size', 'run locally', 'llama.cpp', 'VRAM', or 'smaller models' that users would naturally say.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete actions: '4-bit/8-bit quantization, GGUF conversion, memory optimization, and quality-performance tradeoffs'. These are concrete, technical capabilities.	3 / 3
Completeness	Clearly answers 'what' (quantization, GGUF conversion, memory optimization) but lacks an explicit 'Use when...' clause. The 'when' is only implied through the phrase 'deploying LLMs in resource-constrained JARVIS environments'.	2 / 3
Trigger Term Quality	Includes relevant technical terms like 'quantization', 'GGUF', 'LLMs', but missing common user variations like 'compress model', 'reduce model size', 'llama.cpp', or 'smaller models'. Users may not always use these exact technical terms.	2 / 3
Distinctiveness Conflict Risk	Clear niche focused specifically on model quantization and GGUF conversion. The specific technical domain (AI model optimization for deployment) is distinct and unlikely to conflict with general coding or document skills.	3 / 3
	Total	10 / 12 Passed

Validation

68%

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 — 11 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (552 lines); consider splitting into references/ and linking	Warning
description_trigger_hint	Description may be missing an explicit 'when to use' trigger hint (e.g., 'Use when...')	Warning
metadata_version	'metadata' field is not a dictionary	Warning
license_field	'license' field is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	11 / 16 Passed

Repository: martinholovsky/claude-skills-generator
Commit: 1086ef2

Reviewed: 5 months 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.