deepspeed

Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention

Quality

58%

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 ./skills/distributed-training/deepspeed/SKILL.md

Quality

Content

35%

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

The body is an auto-generated dump: giant verbatim tutorials inline, a malformed code example, and navigation pointing to non-existent files. It is token-heavy and weakly actionable despite real reference files existing on disk.

Suggestions

Replace the inline "Common Patterns" tutorial dumps with a concise quick-reference plus signaled links into references/tutorials.md; remove the duplicated Pattern 3 paragraph.

Fix the malformed ```python block (remove the `### Create aio_handle` line) and align navigation to the actual files (08.md, 09.md, 2020.md, 2023.md, mii.md, other.md, tutorials.md, assets.md, index.md).

Give each reference file a meaningful one-line description instead of "08.md - 08 documentation", and drop the empty scripts/ and assets/ placeholder sections.

Dimension	Reasoning	Score
Conciseness	The "Common Patterns" section pastes entire tutorials verbatim (DeepNVMe, MoE-for-NLG, LRRT) as prose walls, padding the body with documentation Claude could read on demand — the worst-case verbose anchor.	1 / 3
Actionability	A real executable snippet exists (aio_handle creation) but the lone "Example Code Patterns" block is malformed with a leaked markdown heading, and the guidance points to non-existent files, so it is concrete but incomplete/broken.	2 / 3
Workflow Clarity	"Working with This Skill" offers a vague beginner→feature→example path with no validation checkpoints, and its file pointers (getting_started, api, guides) do not exist in the bundle.	2 / 3
Progressive Disclosure	Reference files are listed one level deep, but their labels are meaningless ("08.md - 08 documentation"), navigation points to files that aren't present, and large tutorial content that belongs in references is dumped inline.	2 / 3
	Total	7 / 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.

The description is specific and trigger-rich with a clear DeepSpeed niche, but it omits any explicit "Use when..." trigger guidance, capping completeness. Adding a when-to-use clause would round it out.

Suggestions

Add an explicit trigger clause, e.g. "Use when configuring ZeRO stages, pipeline parallelism, or mixed-precision (FP16/BF16/FP8) training with DeepSpeed."

Lead with a concrete action verb ("Configures", "Optimizes") instead of the vague "Expert guidance for".

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities — "ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention" — rather than vague language, matching the multi-specific-actions anchor.	3 / 3
Completeness	It states what the skill covers but has no "Use when..." clause or equivalent explicit trigger guidance, so per the rubric completeness is capped at 2.	2 / 3
Trigger Term Quality	Covers the natural terms a user would say ("DeepSpeed", "distributed training", "ZeRO", "pipeline parallelism", "FP16/BF16/FP8", "sparse attention") with good breadth of common variations.	3 / 3
Distinctiveness Conflict Risk	DeepSpeed-specific triggers (ZeRO stages, 1-bit Adam, FP8, sparse attention) carve a clear niche 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: OpenLAIR/dr-claw
Path: skills/distributed-training/deepspeed/SKILL.md
Commit: 6da7f7c

Reviewed: about 7 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.