CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

megatron-checkpoint-layout

Bilingual guidance for Megatron checkpoint 1D 2D 3D mp_rank layouts across tensor pipeline and expert parallel dimensions

Quality

72%

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 ./.opencode/skills/megatron-checkpoint-layout/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

67%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 content is well-structured, actionable, and self-contained with concrete patterns and a clear decision procedure, but the bilingual duplication imposes a token-efficiency cost and the workflow lacks explicit inline validation checkpoints.

Suggestions

Consider collapsing the bilingual text into a single primary language with the translation in a referenced file, or clearly justify the duplication to reduce token cost.

Add an explicit validation checkpoint inside the decision procedure (e.g., "Verify the inferred mp_rank pattern exists on disk before recommending a loader").

Move the debugging checklist or expected-outputs into a separate reference file with a one-level-deep link to improve progressive disclosure.

DimensionReasoningScore

Conciseness

Within each language the body is lean and assumes Claude's competence without over-explaining Megatron basics, but the deliberate bilingual duplication roughly doubles the token cost, which the rubric treats as a token-efficiency penalty even though it is intentional.

3 / 5

Actionability

Concrete naming templates (`mp_rank_{tp}`, `mp_rank_{tp}_{pp}`, `mp_rank_{tp}_{pp}_{ep}`) and explicit branch rules ("if expert_parallel_size is passed, require 3D") give mostly executable guidance with only minor gaps.

4 / 5

Workflow Clarity

A numbered 4-step decision procedure, a debugging checklist, and an expected-outputs section provide a clear sequence with verification guidance, though the procedure itself lacks explicit inline validation checkpoints.

4 / 5

Progressive Disclosure

The body is well-organized into clearly headed sections with easy navigation and is self-contained, but there are no file references to split content further and it exceeds the simple-skill under-50-line exception.

4 / 5

Total

15

/

20

Passed

Description

65%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 technically specific and clearly distinct, but it lacks an explicit "when to use" trigger clause and relies on a generic action verb ("guidance for"), which limits its completeness and action specificity.

Suggestions

Add an explicit "Use when..." clause naming concrete scenarios such as diagnosing checkpoint loading errors, designing shard layouts, or converting between TP/PP/EP configurations.

Replace the generic "Bilingual guidance for" lead with concrete action verbs (e.g., "Diagnoses, designs, and converts Megatron checkpoint mp_rank layouts...").

Include common abbreviations and synonyms users say naturally (TP/PP/EP, expert parallel, checkpoint conversion/sharding).

DimensionReasoningScore

Specificity

The description names a concrete domain ("Megatron checkpoint 1D 2D 3D mp_rank layouts across tensor pipeline and expert parallel dimensions") but the only action verb is the generic "guidance for", so it lands at the domain-plus-weak-action level rather than listing several concrete actions.

3 / 5

Completeness

It gives a clear "what" (bilingual guidance on mp_rank layouts) but has no "Use when..." or equivalent explicit "when" trigger clause, which caps completeness at 3 per the rubric guideline.

3 / 5

Trigger Term Quality

It surfaces the natural domain terms a user would say ("Megatron checkpoint", "mp_rank layouts", "tensor pipeline and expert parallel"), though it omits common synonyms/abbreviations like TP/PP/EP or "checkpoint conversion".

4 / 5

Distinctiveness Conflict Risk

The niche is highly specific (Megatron checkpoint mp_rank layout dimensions) with distinct technical triggers and minimal overlap risk with other skills.

5 / 5

Total

15

/

20

Passed

Validation

93%

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

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

Total

15

/

16

Passed

Repository

EvolvingLMMs-Lab/LLaVA-OneVision-2

Path

.opencode/skills/megatron-checkpoint-layout/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.