gguf-quantization

GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.

Quality

77%

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

Fix and improve this skill with Tessl

tessl review fix ./10-optimization/gguf/SKILL.md

Quality

Content

65%

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

Highly actionable content with copy-paste-ready code, but it is long and repetitive, lacks explicit validation feedback loops in its workflows, and keeps substantial advanced material inline rather than offloading it to the existing references. Tightening repetition and moving advanced sections into advanced-usage.md would lift the weaker dimensions.

Suggestions

Consolidate repeated quantize/imatrix commands into one canonical example and cross-reference it from Workflows and Common issues to reduce token cost.

Add an explicit validation checkpoint to the conversion/quantization workflows (e.g., verify the output loads and produces sane output before proceeding, with a fix-and-retry loop), especially for the batch Workflow 3.

Move server mode, streaming, hardware optimization, and tool integrations into references/advanced-usage.md so SKILL.md stays a lean overview with one-level-deep references.

Dimension	Reasoning	Score
Conciseness	The ~417-line body is mostly actionable but repeats the same quantize/imatrix commands across Quick start, Workflow 1, Workflow 3, and Common issues, and re-states Python boilerplate, so it could be tightened rather than earning a 3.	2 / 3
Actionability	Provides fully executable bash and Python examples throughout (convert, quantize, llama-cli, llama-server, llama-cpp-python) that are specific and copy-paste ready.	3 / 3
Workflow Clarity	Workflows are numbered and sequenced, but validation is only a light 'Test' step with no explicit validate->fix->retry loop, and Workflow 3 is a batch operation without verification, which caps workflow clarity at 2.	2 / 3
Progressive Disclosure	The two references (advanced-usage.md, troubleshooting.md) are real and cleanly one-level-deep, but a large amount of advanced material (server mode, streaming, hardware optimization, integrations) lives inline in SKILL.md rather than being split out.	2 / 3
	Total	9 / 12 Passed

Description

90%

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 strong description that concisely states both the capability and explicit use-when triggers, with natural trigger terms and a clear, low-conflict niche. The only gap is that it characterizes scenarios rather than enumerating concrete actions, capping specificity at 2.

Dimension	Reasoning	Score
Specificity	Names the domain ('GGUF format and llama.cpp quantization for efficient CPU/GPU inference') but describes deployment scenarios rather than listing multiple concrete actions, so it is not comprehensive enough for a 3.	2 / 3
Completeness	Clearly answers what it does and provides an explicit 'Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization...' trigger clause, satisfying both what and when.	3 / 3
Trigger Term Quality	Includes natural terms users would actually say ('GGUF', 'llama.cpp', 'quantization', 'consumer hardware', 'Apple Silicon', '2-8 bit') with good coverage and no jargon-only phrasing.	3 / 3
Distinctiveness Conflict Risk	The GGUF + llama.cpp quantization niche with triggers like 'Apple Silicon', 'without GPU requirements', and '2-8 bit' is distinct and unlikely to fire for unrelated skills.	3 / 3
	Total	11 / 12 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: Orchestra-Research/AI-Research-SKILLs
Path: 10-optimization/gguf/SKILL.md
Commit: 773a529

Reviewed: about 13 hours 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.