axolotl

Expert guidance for fine-tuning LLMs with Axolotl - YAML configs, 100+ models, LoRA/QLoRA, DPO/KTO/ORPO/GRPO, multimodal support

Quality

58%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-training/axolotl/SKILL.md

Quality

Content

50%

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

The body is a reasonable quick-reference with some genuinely executable snippets and real reference files, but it mixes them with bare-token code blocks, filler sections, and navigation that points to nonexistent files. Tightening examples and fixing the reference pointers would lift its weakest dimensions.

Suggestions

Replace non-executable code blocks (the bare "context_parallel_size" and "integrations" snippets) with runnable YAML/commands or fold them into the surrounding prose.

Fix progressive disclosure by either creating the referenced getting_started/tutorials files or removing that pointer, and add the existing index.md to the listed references.

Trim boilerplate sections ("Working with This Skill", "Resources", "Notes", "Updating") to the essential update instructions, and add a validation step to any rebuild workflow.

Dimension	Reasoning	Score
Conciseness	Mostly a compact reference, but it carries filler prose ("Comprehensive assistance... generated from official documentation") and restated explanations (Pattern 4's context-parallel arithmetic, Pattern 5's restated benefits) that could be tightened; it is above 1 because it avoids long concept tutorials.	2 / 3
Actionability	Some blocks are executable (FSDP yaml, NCCL test command), but others are bare tokens (Pattern 3 is just "context_parallel_size", Pattern 6 is the word "integrations") or constructor signatures with no usage context, so it falls short of copy-paste-ready completeness.	2 / 3
Workflow Clarity	A minimal sequence exists in "Updating" (re-run scraper, rebuild), but there is no validation or feedback loop anywhere and no real multi-step workflow for the core task; it is above 1 only because a sequence is present.	2 / 3
Progressive Disclosure	The body lists real one-level-deep references (api.md, dataset-formats.md, other.md, all present), but "For Beginners" points to getting_started/tutorials files that do not exist in references/, so navigation is incomplete; it is not 3 because of the broken file pointers.	2 / 3
	Total	8 / 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 is specific and distinctive, naming the tool and several concrete capabilities, but it lacks an explicit trigger clause and uses technical jargon over natural user phrasing. Adding a "Use when..." sentence with common terms would raise its weakest dimensions.

Suggestions

Add an explicit "Use when..." clause, e.g. "Use when fine-tuning or training a model with Axolotl, writing training YAML configs, or applying LoRA/QLoRA/DPO/ORPO methods."

Include natural user phrasings beyond jargon, such as "train a model", "instruction tuning", or "SFT", to improve trigger-term coverage.

Consider trimming the dense method acronym list ("DPO/KTO/ORPO/GRPO") to the most common terms and leaving the rest to the body for conciseness.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities ("fine-tuning LLMs", "YAML configs", "LoRA/QLoRA", "DPO/KTO/ORPO/GRPO", "multimodal support"), matching the comprehensive-actions anchor rather than the partial-coverage anchor below.	3 / 3
Completeness	Clearly states what the skill does but has no "Use when..." trigger clause, which per the guidelines caps completeness at 2; it is above 1 because the capability description is explicit.	2 / 3
Trigger Term Quality	Includes relevant terms ("fine-tuning", "LoRA", "Axolotl") but leans on tool-specific jargon ("DPO/KTO/ORPO/GRPO", "QLoRA") and misses common user phrasings like "train a model" or "SFT", so it is not full coverage.	2 / 3
Distinctiveness Conflict Risk	Names a specific tool ("Axolotl") with a distinct fine-tuning niche, making it unlikely to trigger for the wrong skill, rather than the overlapping-domain anchor at 2.	3 / 3
	Total	10 / 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: synthetic-sciences/openscience
Path: backend/cli/skills/ml-training/axolotl/SKILL.md
Commit: 52845c3

Reviewed: about 14 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.