axolotl

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

Quality

55%

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 ./backend/cli/skills/ml-training/axolotl/SKILL.md

Quality

Content

43%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 grab-bag of doc-scraped snippets with two genuinely useful blocks (FSDP config, NCCL test command) surrounded by bare-identifier 'code blocks', signature-only API stubs, generic boilerplate, and references to files that do not exist. It lacks any coherent fine-tuning workflow and its navigation is undermined by broken pointers.

Suggestions

Add a minimal end-to-end workflow (prepare dataset → write axolotl YAML → launch training → validate checkpoints/eval metrics) so the patterns have a sequence to live in.

Fix the broken navigation: remove or create the nonexistent 'getting_started', 'tutorials', and 'guides' references, and delete the empty scripts/ and assets/ placeholder sections.

Replace bare-identifier code blocks (Patterns 3, 5, 6) with real, complete snippets (e.g. an actual YAML stanza using context_parallel_size, a real save_compressed config), and cut the meta-boilerplate Notes/Updating sections.

Dimension	Reasoning	Score
Conciseness	The patterns are compact and skip concept explanations, but placeholder boilerplate ('Add helper scripts here for common automation tasks', the Notes/Updating sections, 'Comprehensive assistance... generated from official documentation') and mangled prose in Patterns 4 and 6 (inline lists crammed into paragraphs) add tokens without value.	3 / 5
Actionability	Pattern 2's FSDP YAML block and Pattern 1's NCCL command are concrete and executable, but Patterns 3, 5, 6 and 7 have code blocks containing only a bare identifier ('context_parallel_size', 'integrations'), and the five API examples are bare signatures (e.g. 'cli.cloud.modal_.ModalCloud(config, app=None)') with no imports or calling context — concrete guidance that is incomplete rather than copy-paste ready.	3 / 5
Workflow Clarity	No multi-step fine-tuning workflow exists anywhere in the body — nothing sequences dataset preparation, YAML config authoring, launch, or evaluation, and there are no validation checkpoints for long-running training jobs. The trigger list under 'When to Use This Skill' is generic boilerplate ('Working with axolotl', 'Asking about axolotl features'), not a sequence; at best there is a rough when-to-use → patterns → references organization with many gaps.	2 / 5
Progressive Disclosure	The 'Reference Files' section cleanly signals the one-level-deep bundle (api.md, dataset-formats.md, other.md, all of which exist with brief descriptions), but the body also directs readers to nonexistent files ('Start with the getting_started or tutorials reference files', 'the appropriate category reference file (api, guides, etc.)') and to empty scripts/ and assets/ sections, so navigation is only partially reliable.	3 / 5
	Total	11 / 20 Passed

Description

66%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 specific, keyword-rich description for a well-defined niche, weakened only by the complete absence of an explicit 'when to use' trigger clause and a slightly marketing-toned opener ('Expert guidance', '100+ models').

Suggestions

Append an explicit trigger clause, e.g. 'Use when fine-tuning or training LLMs with Axolotl, writing Axolotl YAML configs, or applying LoRA/QLoRA/DPO/GRPO.'

Add natural synonyms users would say ('train a model', 'PEFT', 'SFT', 'RLHF', '.yaml') to broaden keyword coverage.

Replace the marketing phrase 'Expert guidance' and '100+ models' with a stated action like 'Configure and run fine-tuning jobs with Axolotl'.

Dimension	Reasoning	Score
Specificity	The description lists several concrete capability areas — 'YAML configs, 100+ models, LoRA/QLoRA, DPO/KTO/ORPO/GRPO, multimodal support' — with only minor gaps, though 'Expert guidance' is generic and the items are topics rather than stated actions, keeping it below a 5.	4 / 5
Completeness	The 'what' is clear ('Expert guidance for fine-tuning LLMs with Axolotl'), but there is no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 3 per the judging guidelines.	3 / 5
Trigger Term Quality	Natural terms users would say are well covered ('fine-tuning LLMs', 'LoRA', 'QLoRA', 'DPO', 'GRPO', 'YAML', 'multimodal'), but common synonyms like 'train/training', 'PEFT', 'SFT/RLHF', or '.yaml' files are missing, so it is not comprehensive.	4 / 5
Distinctiveness Conflict Risk	'Axolotl' pins a clear niche with distinct triggers, but the generic fine-tuning vocabulary (LoRA, DPO, multimodal) creates minor overlap risk with closely related fine-tuning skills, so it is not a 5.	4 / 5
	Total	15 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/ml-training/axolotl/SKILL.md
Commit: 4082a2e

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.