CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

distributed-training-ray-train

Distributed training orchestration across clusters. Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nodes. Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling. Use when training massive models across multiple machines or running distributed hyperparameter sweeps.

Quality

79%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./cli-tool/components/skills/ai-research/distributed-training-ray-train/SKILL.md
SKILL.md
Quality
Evals
Security

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.

A well-organized, largely executable skill body with good use-case coverage and useful troubleshooting, held back by marketing padding and generic explanations, partly non-runnable example snippets, and two of its three referenced files missing from the bundle.

Suggestions

Add the missing references/hyperparameter-tuning.md and references/custom-loops.md files (or remove the links) so all 'Advanced topics' references resolve — this is the single biggest progressive-disclosure defect.

Trim marketing fluff and time-sensitive details from the Resources/Hardware sections (star counts, 'Used by' customer list, 'Version: 2.40.0+') to improve token efficiency.

Make example snippets self-contained: define or clearly stub MyModel/train_epoch/train_dataset, and correct the Workflow 4 checkpoint save (checkpoint.path after Checkpoint.from_directory) so examples run as written.

DimensionReasoningScore

Conciseness

Mostly efficient, code-first content, but padded with marketing fluff ('Best-in-class hyperparameter tuning', 'Used by: OpenAI, Uber, Spotify, Shopify, Instacart', 'GitHub ... ⭐ 36,000+'), a version pin ('Version: 2.40.0+') outside any deprecated/old-patterns section, and generic explanations Claude already knows (gradient accumulation, timing a dataloader loop). Not 4 because these unnecessary sections are noticeable rather than minor.

3 / 5

Actionability

Mostly executable, copy-paste-ready examples for the common cases (quick start, Tune search space, cluster launch commands), with minor gaps: snippets depend on undefined placeholders (MyModel, train_epoch, train_dataset, dataloader), and the Workflow 4 checkpoint code is questionable (saving via checkpoint.path after Checkpoint.from_directory of the trial dir). Not 5 because several workflows would not run as written.

4 / 5

Workflow Clarity

Use-case workflows are clearly presented (install → define train_func → configure ScalingConfig → fit) and the 'Common issues' section provides check-then-fix feedback loops (ray status → restart head/workers; reduce workers for OOM). Not 5 because there are no explicit validation checkpoints before launching expensive multi-node jobs (e.g., verifying cluster connectivity or GPU availability up front).

4 / 5

Progressive Disclosure

The body is well-sectioned and the 'Advanced topics' section clearly signals references, but 2 of the 3 referenced files (references/hyperparameter-tuning.md and references/custom-loops.md) do not exist in the bundle — only references/multi-node.md is present. Scored against the actual bundle structure, broken references are more than a minor organization gap; not 4 because following two of the three links yields nothing, undermining navigation.

3 / 5

Total

14

/

20

Passed

Description

88%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 strong description: concrete capabilities, explicit 'Use when...' triggers covering both multi-machine training and distributed hyperparameter sweeps, and a clear Ray Train niche. Main improvement is broadening natural trigger synonyms (multi-node, cluster, multi-GPU) and tightening distinctiveness versus adjacent Ray/Tune skills.

DimensionReasoningScore

Specificity

Lists multiple concrete capabilities — 'Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nodes', 'Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling' — giving comprehensive, specific coverage of the skill's actions rather than vague domain naming.

5 / 5

Completeness

Explicitly answers both: what ('Distributed training orchestration across clusters... hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling') and when ('Use when training massive models across multiple machines or running distributed hyperparameter sweeps') with concrete trigger phrases.

5 / 5

Trigger Term Quality

Good natural keyword coverage ('training massive models across multiple machines', 'distributed hyperparameter sweeps', framework names), but common variations like 'multi-node', 'cluster', or 'multi-GPU' training are missing. Not score 5 because the synonym/file-extension-level breadth of the top anchor is absent; not score 3 because it goes well beyond 'some relevant keywords'.

4 / 5

Distinctiveness Conflict Risk

Clear niche in Ray-based distributed/multi-node training with distinct triggers, but minor overlap risk with closely related skills such as a general Ray, Ray Tune, or PyTorch distributed training skill. Not 5 because 'training massive models' is a trigger a general ML-training skill could also claim.

4 / 5

Total

18

/

20

Passed

Validation

81%

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

CriteriaDescriptionResult

metadata_field

'metadata' should map string keys to string values

Warning

relative_links

Relative link issues: 2 missing

Warning

referenced_paths_exist

Referenced path issues: 4 missing

Warning

Total

13

/

16

Passed

Repository

davila7/claude-code-templates

Path

cli-tool/components/skills/ai-research/distributed-training-ray-train/SKILL.md

Reviewed

Table of Contents

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.