llama-factory

Expert guidance for fine-tuning LLMs with LLaMA-Factory - WebUI no-code, 100+ models, 2/3/4/5/6/8-bit QLoRA, multimodal support

Quality

40%

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 ./03-fine-tuning/llama-factory/SKILL.md

Quality

Content

22%

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

The body is an auto-generated scaffold: an empty Quick Reference placeholder, no executable guidance, no real fine-tuning workflow, and generic reference descriptions with some dangling file references. It functions as a directory listing rather than actionable skill content.

Suggestions

Replace the empty 'Quick reference patterns will be added' placeholder with concrete, copy-paste-ready examples (e.g., a minimal WebUI QLoRA training command and a YAML config snippet).

Add a sequenced fine-tuning workflow with explicit validation checkpoints (prepare dataset -> configure YAML -> launch training -> verify loss/eval -> merge/export adapters).

Fix the dangling references (remove or create 'api', 'guides', 'tutorials') and replace generic file descriptions with one-line summaries of what each reference file actually contains.

Dimension	Reasoning	Score
Conciseness	The body is short rather than verbose, but it is padded with non-actionable boilerplate (the 'Notes', generic 'Resources' descriptions, and an empty 'Quick reference patterns will be added' placeholder) that does not earn its place and could be tightened considerably.	2 / 3
Actionability	There is no concrete code or command anywhere in the body, and the Quick Reference section is an explicit empty placeholder ('Quick reference patterns will be added as you use the skill'), so it describes rather than instructs.	1 / 3
Workflow Clarity	No sequenced workflow for the actual fine-tuning task (install, prepare data, configure, train, evaluate) is provided, and there are no validation checkpoints for the risky/batch training operations the skill covers; the only numbered steps are for refreshing the skill itself.	1 / 3
Progressive Disclosure	Reference files are real and one level deep with a listed index, but their descriptions are generic ('Advanced documentation') rather than clearly signaling content, and the body points to category files (api, guides, tutorials) that do not exist in the bundle.	2 / 3
	Total	6 / 12 Passed

Description

57%

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 specific and distinctive thanks to the named tool and feature list, but it lacks any explicit 'Use when...' trigger guidance and leans on feature-catalog phrasing rather than concrete action verbs. It is solid but generic at the margins.

Suggestions

Add an explicit 'Use when...' clause naming the natural trigger situations (e.g., 'Use when fine-tuning or quantizing LLMs with LLaMA-Factory, configuring WebUI training, or setting up QLoRA/LoRA adapters').

Lead with concrete action verbs (Train, Fine-tune, Quantize, Merge adapters) instead of 'Expert guidance for', and drop the bit-width enumeration jargon in favor of natural terms like 'LoRA' and 'quantize'.

Include common trigger variations users actually say (LoRA, train a model, quantize, inference) to broaden trigger coverage.

Dimension	Reasoning	Score
Specificity	Names the domain ('fine-tuning LLMs with LLaMA-Factory') and several features (WebUI no-code, QLoRA, multimodal), but leads with the abstract phrase 'Expert guidance for' rather than listing multiple concrete actions, so it does not reach the level-3 bar of enumerated actions.	2 / 3
Completeness	Clearly states what the skill covers, but provides no 'Use when...' clause or equivalent explicit trigger guidance, so per the rubric guideline completeness is capped at 2 (the 'when' is entirely missing).	2 / 3
Trigger Term Quality	Includes relevant natural terms (fine-tuning, LLaMA-Factory, WebUI, QLoRA, multimodal) but is missing common variations users would say (LoRA, train, quantize, inference), and the '2/3/4/5/6/8-bit' enumeration reads as technical jargon padding rather than natural speech.	2 / 3
Distinctiveness Conflict Risk	Anchoring to the specific named tool 'LLaMA-Factory' alongside QLoRA/WebUI/multimodal carves a clear niche with distinct triggers that is unlikely to fire for an unrelated skill.	3 / 3
	Total	9 / 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: 03-fine-tuning/llama-factory/SKILL.md
Commit: 773a529

Reviewed: about 12 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.