unsloth

Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization

Quality

41%

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 ./bundled/skills/unsloth/SKILL.md

Quality

Content

21%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.

The body is auto-generated boilerplate with no executable guidance, no concrete workflow, and references to reference categories (api, guides, tutorials, getting_started) that do not exist; it relies on a single 813KB monolithic reference file with no navigable structure.

Suggestions

Replace placeholder boilerplate ('Quick reference patterns will be added...', 'Add helper scripts here', 'This skill was automatically generated') with concrete, executable Unsloth snippets (model loading, LoRA config, training call).

Point to the actual reference files that exist (references/llms-txt.md, llms.md) instead of non-existent categories like api/, guides/, tutorials/, getting_started.

Add a real quick-start workflow with explicit steps and a validation checkpoint (e.g., run a short training step and confirm loss decreases before full runs).

Dimension	Reasoning	Score
Conciseness	The body is padded with generic boilerplate that adds nothing Claude doesn't already know ('Add helper scripts here', 'Add templates, boilerplate, or example projects here', 'Quick reference patterns will be added as you use the skill') and meta-commentary about how the skill was generated; this is noticeably verbose and padded without delivering substance, matching the score-2 anchor.	2 / 5
Actionability	There is no executable code, no concrete commands, and no specific guidance anywhere in the body — only abstract pointers like 'Use the appropriate category reference file (api, guides, etc.)' for categories that do not exist in references/; this is entirely vague and descriptive, matching the score-1 anchor.	1 / 5
Workflow Clarity	A rough 'When to Use' list and a vague reference to reading reference files give a coarse sequence with no concrete steps, no commands, and no validation checkpoints; it has rough structure but major gaps, matching the score-2 anchor.	2 / 5
Progressive Disclosure	There is structure with section headers and a one-level reference to references/llms-txt.md (which exists), but navigation is weakly signaled: the body mentions nonexistent categories (getting_started, tutorials, api, guides) that are not present in references/, and the single huge 813KB llms-txt.md (136 pages) is not broken into navigable sub-files, so structure is present but poorly organized — a 3.	3 / 5
	Total	8 / 20 Passed

Description

61%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.

The description is concrete and domain-specific with good natural keywords, but it lacks an explicit 'Use when...' trigger clause and frames capabilities as performance metrics rather than concrete actions, capping completeness and specificity.

Suggestions

Add an explicit trigger clause, e.g. 'Use when fine-tuning LLMs (Llama, Mistral, Gemma, Qwen) with Unsloth for faster, memory-efficient training.'

Replace performance-marketing phrasing ('2-5x faster', '50-80% less memory') with concrete actions like 'configure LoRA/QLoRA adapters, prepare datasets, and run memory-efficient training jobs'.

Include common synonyms and concrete extensions (e.g. '.safetensors', adapter checkpoints) to broaden trigger-term coverage.

Dimension	Reasoning	Score
Specificity	It names the domain (Unsloth fine-tuning) and the concrete optimization outcomes ('2-5x faster training', '50-80% less memory', 'LoRA/QLoRA optimization'), which matches the 'names domain and 1-2 concrete actions' anchor; however, it lists performance metrics rather than concrete actions (e.g., 'prepare datasets', 'configure LoRA'), so it stops short of a 4.	3 / 5
Completeness	It clearly answers 'what' (fast fine-tuning guidance with LoRA/QLoRA optimization) but has no explicit 'Use when...' / 'when should Claude use it' clause, which the guidelines say caps completeness at 3.	3 / 5
Trigger Term Quality	'fine-tuning', 'LoRA/QLoRA', and 'training' are natural terms users say in this domain, and 'Unsloth' plus 'memory'/'faster training' give good keyword coverage; it falls just short of comprehensive coverage of synonyms and concrete file/format extensions, so a 4 rather than 5.	4 / 5
Distinctiveness Conflict Risk	The Unsloth-specific framing plus LoRA/QLoRA keywords carve a fairly distinct niche with only minor overlap risk against generic fine-tuning/PEFT skills; it is not a fully conflict-free niche (could overlap with general PEFT fine-tuning skills), so 4 rather than 5.	4 / 5
	Total	14 / 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: foryourhealth111-pixel/Vibe-Skills
Path: bundled/skills/unsloth/SKILL.md
Commit: ddcaa2a

Reviewed: 6 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.