pytorch-lightning

High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.

Quality

76%

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 ./08-distributed-training/pytorch-lightning/SKILL.md

Quality

Content

70%

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

A well-structured, highly actionable skill body with excellent progressive disclosure and executable examples. Main weakness is repetition of the same LightningModule across multiple workflows and implicit rather than explicit validation checkpoints for risky training operations.

Suggestions

De-duplicate the repeated LitModel definitions across Workflows 1-5; define the canonical module once and show only the delta in subsequent workflows to reduce tokens.

Add an explicit validation feedback loop (e.g., check val_loss is decreasing, else reduce lr or check data) rather than relying on implicit 'automatic' checkpointing notes.

Trim promotional filler ('That's it!', 'battle-tested', '1M+ downloads/month') that does not guide behavior.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient with executable code, but it includes redundant re-derivations (e.g., the same LitModel redefined across Workflow 1, 2, 3, 5), restated 'Automatic features' bullets, and mildly promotional lines like 'That's it!' and 'battle-tested', which could be tightened.	2 / 3
Actionability	Provides fully executable, copy-paste-ready code across installation, conversion, validation/testing, distributed training, callbacks, scheduling, and troubleshooting, with specific Trainer flags and concrete fixes for each issue.	3 / 3
Workflow Clarity	Workflows are clearly sequenced and numbered (the 3-step conversion is crisp), but training/validation operations involve checkpointing and distributed runs where validation checkpoints are described as automatic rather than as explicit verify-fix-retry steps, so feedback loops are implicit.	2 / 3
Progressive Disclosure	The body is a concise overview that points one level deep to real bundle files (references/callbacks.md, distributed.md, hyperparameter-tuning.md), each of which exists and is clearly signaled under 'Advanced topics' with descriptive labels.	3 / 3
	Total	10 / 12 Passed

Description

82%

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 strong description: third-person, specific, with an explicit 'Use when' clause covering both capability and trigger. It could improve trigger-term quality by adding more natural phrases a user might say rather than relying on framework jargon.

Suggestions

Add natural user-facing trigger phrases (e.g., 'train a PyTorch model', 'multi-GPU training', 'scale training across GPUs') alongside the technical DDP/FSDP/DeepSpeed terms.

Consider mentioning a couple of plain-language pain points the skill solves (e.g., 'when boilerplate training loops are getting in the way') to broaden natural trigger coverage.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions and capabilities: 'Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate', matching the anchor that enumerates several specific concrete actions.	3 / 3
Completeness	It states what the skill does (high-level framework with Trainer, distributed training, callbacks) and explicitly answers when to use it via 'Use when you want clean training loops with built-in best practices', satisfying both the 'what' and the explicit 'Use when...' trigger.	3 / 3
Trigger Term Quality	Contains relevant terms like 'PyTorch', 'training loops', 'distributed training', and 'DDP/FSDP/DeepSpeed', but leans on technical jargon rather than natural user phrases; it omits common variations a user would naturally say such as 'train a model', 'GPU training', or 'fast training'.	2 / 3
Distinctiveness Conflict Risk	The PyTorch Lightning niche is clearly delineated by framework-specific triggers (Trainer, DDP/FSDP/DeepSpeed, callbacks) making it unlikely to fire for unrelated skills.	3 / 3
	Total	11 / 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: 08-distributed-training/pytorch-lightning/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.