pytorch-training

Best practices for building robust PyTorch training loops. Use when generating or reviewing ML training code.

Quality

71%

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 ./researchclaw/skills/builtin/tooling/pytorch-training/SKILL.md

Quality

Content

80%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 lean and token-efficient with specific, actionable API references and good structure for a simple skill. Its main gap is the absence of a sequenced training workflow with validation checkpoints.

Suggestions

Reorder the list into a sequenced training workflow (setup -> training loop -> validation/early stopping -> checkpointing) with explicit validation checkpoints between stages.

Add a short executable training-loop snippet in the body (or surface the frontmatter code template) so guidance is copy-paste ready rather than directive-only.

Add an explicit validation step, e.g. 'Evaluate on validation set each epoch with torch.no_grad() before checkpointing the best model.'

Dimension	Reasoning	Score
Conciseness	An 8-line best-practice list with no padding or over-explanation of concepts Claude already knows; every line earns its place.	5 / 5
Actionability	Names specific APIs (e.g. clip_grad_norm_, set_to_none=True, CosineAnnealingLR, torch.no_grad()) but the body is directive-style without an executable code block in the content itself.	4 / 5
Workflow Clarity	The body is an unordered checklist rather than a sequenced training workflow, and lacks validation checkpoints for a batch/iterative operation, which caps the score at 3.	3 / 5
Progressive Disclosure	Under 50 lines, single-purpose, and well-organized into one clean list with no need for external references, qualifying for the simple-skill exception.	5 / 5
	Total	17 / 20 Passed

Description

62%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 concise, third-person, and answers both what and when, with a clear PyTorch-training niche. Its main weakness is generic trigger phrasing and limited breadth of named actions.

Suggestions

Expand the 'Use when' clause with concrete trigger phrases users would naturally say, e.g. 'training loops, model training, fine-tuning, or writing PyTorch training code'.

List a few more specific capabilities (e.g. checkpointing, learning-rate scheduling, distributed training) to raise specificity beyond two actions.

Dimension	Reasoning	Score
Specificity	Names the PyTorch training domain and two concrete actions ('generating or reviewing ML training code'), but offers only a couple of actions rather than a comprehensive list.	3 / 5
Completeness	Clearly states what ('building robust PyTorch training loops') and an explicit when ('Use when generating or reviewing ML training code'), though the trigger phrasing is generic rather than concrete.	4 / 5
Trigger Term Quality	'generating or reviewing ML training code' provides some relevant keywords but misses common natural variations like 'training loops', 'model training', or 'fine-tuning' that users would actually say.	3 / 5
Distinctiveness Conflict Risk	The PyTorch training niche is mostly distinct with clear triggers, with only minor overlap risk against broader deep-learning or ML skills.	4 / 5
	Total	14 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: aiming-lab/AutoResearchClaw
Path: researchclaw/skills/builtin/tooling/pytorch-training/SKILL.md
Commit: be4ba47

Reviewed: 5 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.