length-pool-sort-dataset

Bilingual guide for understanding LengthPoolSortDataset cross-rank length synchronization mechanism in multi-GPU training

Quality

53%

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 ./.opencode/skills/length-pool-sort-dataset/SKILL.md

Quality

Content

50%

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

The body is a thorough, well-structured explanation of the LengthPoolSortDataset mechanism with concrete code and checklists, but it pays a large token cost from full bilingual duplication and re-explaining known concepts, and it is monolithic with no progressive disclosure. Splitting deep-dive material into reference files and trimming redundant bilingual lines would improve efficiency.

Suggestions

Move the detailed step-by-step "why it works" deep-dive and tuning tables into a reference file, keeping SKILL.md as a concise overview with a clearly signaled link.

Reduce bilingual duplication by leading with one language and offering the other as a referenced section, or condensing parallel translations.

Add an explicit validation checkpoint to the debugging workflow (e.g., "confirm step-time variance dropped before declaring the bottleneck resolved").

Dimension	Reasoning	Score
Conciseness	Content is informative and not fluffy, but every section is duplicated in Chinese (roughly doubling tokens) and it re-explains basic distributed-training concepts (all-reduce, data-parallel) Claude already knows.	2 / 3
Actionability	Provides a concrete code snippet, the --length-sort-pool-size flag, a tuning table, and a debugging checklist, but the code references class internals (self.dataset, self.key_fn) and is illustrative rather than standalone-executable.	2 / 3
Workflow Clarity	The three-step mechanism and debugging checklist are clearly laid out, but there are no explicit validation/verification checkpoints or feedback loops for the analysis process.	2 / 3
Progressive Disclosure	Sections are well-organized with clear headers, but the skill is a monolithic 200+ line SKILL.md with no bundle files and no offloaded references that could be split out.	2 / 3
	Total	8 / 12 Passed

Description

57%

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 to a well-defined niche and unlikely to conflict with other skills, but it lacks an explicit "Use when…" trigger and reads as jargon-heavy rather than natural-language guidance. Adding trigger phrasing and one or two more concrete actions would lift completeness and trigger quality.

Suggestions

Add a "Use when…" clause naming natural triggers (e.g., "Use when debugging slow multi-GPU training caused by length variance, or when tuning --length-sort-pool-size").

Replace or supplement "guide for understanding" with concrete verbs users would say (analyze, debug, tune) to broaden trigger-term coverage.

Keep the mechanism name but pair it with a plainer-language summary so the description reads less like internal jargon.

Dimension	Reasoning	Score
Specificity	Names the concrete domain (multi-GPU training) and the specific mechanism (LengthPoolSortDataset, cross-rank length synchronization), but only a single action ("understanding") rather than multiple distinct concrete actions.	2 / 3
Completeness	Clearly states what the skill does ("guide for understanding LengthPoolSortDataset cross-rank length synchronization mechanism") but provides no "Use when…" trigger clause, capping completeness at 2.	2 / 3
Trigger Term Quality	Terms like "LengthPoolSortDataset" and "multi-GPU training" are relevant and would be said by users of this repo, but the phrasing is jargon-heavy with no natural trigger variations.	2 / 3
Distinctiveness Conflict Risk	Targets a highly specific named mechanism in a specific repo, making it clearly distinguishable and unlikely to trigger for the wrong skill.	3 / 3
	Total	9 / 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
metadata_version	'metadata.version' is missing	Warning

	Total	15 / 16 Passed

Repository: EvolvingLMMs-Lab/LLaVA-OneVision-2
Path: .opencode/skills/length-pool-sort-dataset/SKILL.md
Commit: bad6f1d

Reviewed: 4 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.