distributed-training

Multi-GPU and distributed training patterns with PyTorch DDP. Use when scaling training across GPUs.

Quality

73%

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/distributed-training/SKILL.md

Quality

Content

72%Weight 40%Scale 1-3

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

A token-efficient, well-organized checklist of concrete distributed-training practices. It scores lower on actionability and workflow clarity because several steps omit complete executable code and no validation or sequencing-as-a-process is provided.

Suggestions

Add brief executable snippets or formulas for the under-specified steps, e.g. 'sampler = DistributedSampler(dataset)' and 'lr = base_lr * world_size'.

Reframe the list as an ordered workflow with a validation checkpoint, e.g. 'Verify all ranks report the same loss before saving the checkpoint on rank 0'.

Dimension	Reasoning	Score
Conciseness	The body is a lean list of short imperative directives with concrete API calls and no padding or explanation of concepts Claude already knows, matching the 'lean and efficient; every token earns its place' anchor.	3 / 3
Actionability	Some items give executable code ('dist.init_process_group(backend='nccl')', 'nn.SyncBatchNorm.convert_sync_batchnorm()'), but others like 'Use DistributedSampler for data sharding' and 'Scale learning rate linearly with world size' lack complete code or formulas, fitting the 'some concrete guidance but incomplete' anchor rather than the copy-paste-ready level 3.	2 / 3
Workflow Clarity	The numbered list provides a sequence of best practices but lacks validation checkpoints or error-recovery feedback loops, matching the 'steps listed but validation gaps' anchor; it reads as a checklist rather than a coherent validated workflow.	2 / 3
Progressive Disclosure	No bundle files exist and the skill is well under 50 lines with a single well-organized section, so per the rubric's simple-skills guideline it qualifies for the top score with just well-organized sections.	3 / 3
	Total	10 / 12 Passed

Description

75%Weight 40%Scale 1-3

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 clear, concise description with explicit what and when clauses and a distinct niche. It is held back from full marks only by limited action specificity and trigger-term variation.

Suggestions

Replace the generic word 'patterns' with concrete actions, e.g., 'Initialize process groups, shard data with DistributedSampler, and sync batch norm for multi-GPU training'.

Broaden trigger terms to include natural variations users say, such as 'multi-GPU', 'data-parallel', 'scale training', or 'distributed training'.

Dimension	Reasoning	Score
Specificity	Phrases like 'distributed training patterns with PyTorch DDP' name the domain and a specific framework, but 'patterns' is not a list of multiple concrete actions, matching the anchor for 'names domain and some actions, but not comprehensive' rather than the multi-action level 3 anchor.	2 / 3
Completeness	The description states both what it does ('distributed training patterns with PyTorch DDP') and an explicit when clause ('Use when scaling training across GPUs'), clearly answering both questions, matching the level 3 anchor.	3 / 3
Trigger Term Quality	'Use when scaling training across GPUs' plus 'multi-GPU' and 'DDP' provide relevant natural keywords a user would say, but coverage of common variations is limited, fitting the 'some relevant keywords but missing common variations' anchor rather than the broad level 3 coverage.	2 / 3
Distinctiveness Conflict Risk	The PyTorch DDP / distributed training niche with its explicit trigger is clearly distinguishable and unlikely to trigger for unrelated skills, matching the 'clear niche with distinct triggers' anchor.	3 / 3
	Total	10 / 12 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/distributed-training/SKILL.md
Commit: e2e23c9

Reviewed: 6 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.