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

75%

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 ./backend/cli/skills/ml-inference/gguf/SKILL.md

Quality

Content

68%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 highly actionable, code-rich skill body with clear section structure and valid one-level references, weakened mainly by redundancy between the Quick start and Conversion workflows sections and a batch workflow lacking a validation checkpoint.

Suggestions

Consolidate the Quick start and Conversion workflows sections to eliminate the duplicated convert+quantize steps, or have Quick start link into the workflows instead of repeating them.

Add an explicit validation/test checkpoint to the batch Workflow 3 (e.g., verify each output loads with ./llama-cli before reporting success) so destructive/batch operations have a feedback loop.

Move the Python usage, server mode, and hardware optimization sections into reference files (similar to advanced-usage.md) so SKILL.md stays a lean overview.

Dimension	Reasoning	Score
Conciseness	Mostly actionable content, but it opens by explaining what GGUF is ('The GGUF (GPT-Generated Unified Format) is the standard file format for llama.cpp...') and repeats material: the Quick start covers convert/quantize/run, then the Conversion workflows section re-covers the same convert+quantize steps in workflows 1-3.	3 / 5
Actionability	Fully executable, copy-paste-ready bash and python covering installation, conversion, quantization, inference, server mode, hardware optimization, and tool integrations, with a concrete quantization-type reference table.	5 / 5
Workflow Clarity	Workflows are clearly numbered, but Workflow 3 (multiple quantizations) is a batch operation with no validation/verification checkpoint before proceeding, which per the rubric caps workflow clarity at 3.	3 / 5
Progressive Disclosure	Well-organized with clear section headers and two real one-level-deep references (references/advanced-usage.md, references/troubleshooting.md) that both exist as files; however, most content (quantization tables, multiple workflows, Python examples, hardware tuning, integrations) is inlined in a ~428-line SKILL.md rather than split out.	4 / 5
	Total	15 / 20 Passed

Description

82%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 strong, third-person description that clearly states what the skill does and when to use it with concrete, natural triggers. Its only gap is specificity, as it describes the domain rather than enumerating the concrete operations (convert, quantize, run inference).

Dimension	Reasoning	Score
Specificity	Names the domain ('GGUF format and llama.cpp quantization') and 1-2 concrete actions ('quantization', 'CPU/GPU inference') but does not enumerate the full action set (convert, quantize, run), so it is not comprehensive.	3 / 5
Completeness	Explicitly answers both 'what' ('GGUF format and llama.cpp quantization for efficient CPU/GPU inference') and 'when' ('Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Good coverage of natural terms users say ('GGUF', 'llama.cpp', 'Apple Silicon', 'consumer hardware', 'quantization 2-8 bit'), though it omits the .gguf file extension and a few synonyms.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (GGUF/llama.cpp CPU-Apple-Silicon inference) with distinct triggers and minimal overlap risk; it even names alternative formats (AWQ/GPTQ, HQQ) as out-of-scope.	5 / 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: synthetic-sciences/openscience
Path: backend/cli/skills/ml-inference/gguf/SKILL.md
Commit: 3a6c3a9

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.