Name: hugging-face-model-trainer
Rating: 52 (1 reviews)
Author: administrakt0r

hugging-face-model-trainer

Train language models using TRL (Transformer Reinforcement Learning) on fully managed Hugging Face infrastructure. No local GPU setup required—models train on cloud GPUs and results are automatically saved to the Hugging Face Hub.

Quality

58%

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 ./plugins/AI-Agents-Safe-Coding-Skills-claude/skills/hugging-face-model-trainer/SKILL.md

The canonical home for this skill is hugging-face-model-trainer in administrakt0r/AI-Agents-Safe-Coding-Skills

Quality

Content

63%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 content is highly actionable with a clear, validated workflow, but it suffers from notable cross-section repetition and references a bundle (references/ and scripts/) that does not exist, breaking progressive disclosure. Consolidating duplicated warnings and shipping the referenced files would materially improve it.

Suggestions

Create the referenced bundle files (references/training_methods.md, scripts/train_sft_example.py, etc.) or remove the references — currently every 'See references/X.md' link is broken since no references/ or scripts/ directory exists.

Consolidate the repeated timeout and Hub-push warnings into a single authoritative section and cross-reference it instead of restating them in Prerequisites, Timeout Management, Common Failure Modes, Troubleshooting, and Key Takeaways.

Move the large inline hardware-selection table, timeout table, and GGUF conversion detail into their referenced reference files so the SKILL.md body stays a lean overview.

Dimension	Reasoning	Score
Conciseness	The body is largely actionable rather than concept-explaining, but the same warnings are repeated across many sections (the 30-min timeout warning appears ~5 times, the ephemeral-env/Hub-push warning ~4 times), which is padding that could be tightened.	3 / 5
Actionability	It provides concrete, mostly copy-paste-ready code blocks and CLI commands for the common cases (hf_jobs UV submission, CLI syntax, configs), with only minor gaps such as the '<see references/gguf_conversion.md for complete script>' placeholder.	4 / 5
Workflow Clarity	The asynchronous-job workflow is clearly sequenced (prerequisites -> create script -> submit -> report -> wait) with validation checkpoints (dataset validation, pre-submit verification checklist) and a failure-mode/fix section, though a few checkpoints are implicit rather than enforced.	4 / 5
Progressive Disclosure	Section structure is well-organized and references are clearly signaled ('See references/X.md'), but every referenced bundle file (references/.md, scripts/.py) is absent — no references/, scripts/, or assets/ directories exist — so navigation is broken and inlined content that belongs in those files cannot be offloaded.	3 / 5
	Total	14 / 20 Passed

Description

53%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 clearly states what the skill does but lacks any explicit 'when to use' trigger guidance, capping completeness. Trigger-term coverage is moderate and leans on the technical acronym TRL rather than user-natural phrases.

Suggestions

Add a 'Use when...' clause naming natural trigger phrases, e.g. 'Use when the user wants to fine-tune or train a language model on cloud GPUs without local setup.'

Replace or augment 'TRL' with user-natural synonyms like 'fine-tune', 'fine-tuning', and 'LLM training' so the description matches how users actually phrase the request.

List a few more concrete capabilities (e.g. SFT/DPO/GRPO methods, GGUF conversion for local deployment) to lift specificity from 1-2 actions toward comprehensive coverage.

Dimension	Reasoning	Score
Specificity	It names the domain ("Train language models using TRL") and a couple concrete actions (train on cloud GPUs, auto-save to the Hub) but coverage is not comprehensive — no mention of fine-tuning, conversion, or the distinct training methods.	3 / 5
Completeness	The 'what' is clear (train LMs on managed HF infra with auto-save), but there is no 'Use when...' clause or equivalent trigger guidance, which per the rubric caps completeness at 3.	3 / 5
Trigger Term Quality	It includes some natural terms ("train language models", "Hugging Face") but relies on the jargon acronym "TRL" and omits common synonyms users would actually say like "fine-tune", "fine-tuning", or "LLM".	3 / 5
Distinctiveness Conflict Risk	The combination of TRL plus fully-managed Hugging Face Jobs infrastructure is a fairly distinct niche with minimal overlap risk, though it could still collide with a generic fine-tuning skill.	4 / 5
	Total	13 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (721 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 35 missing	Warning

	Total	13 / 16 Passed

Repository: administrakt0r/AI-Agents-Safe-Coding-Skills
Path: plugins/AI-Agents-Safe-Coding-Skills-claude/skills/hugging-face-model-trainer/SKILL.md
Commit: 57c135b

Reviewed: 4 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.