llava-onevision2-consistency

Bilingual guide for running and interpreting LLaVA-OneVision2 HF vs Megatron consistency checks across TP and PP settings

Quality

75%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./.opencode/skills/llava-onevision2-consistency/SKILL.md

Quality

Content

77%

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

The body is highly actionable with clear, validated workflows, but bilingual line-by-line duplication inflates token cost and the single-file monolith misses an opportunity to split large reference material into separate files.

Suggestions

Eliminate the bilingual duplication or move the Chinese text into a separate localized reference file to recover roughly half the token budget for single-language use.

Split the large reference-heavy sections (e.g., the 6-test specs, weight-mapping table, fixture list) into dedicated reference files with one-level-deep links from SKILL.md to improve progressive disclosure.

Place the time-sensitive 'verified 2026-05-25' round-trip correctness note in a dedicated validation/verification subsection so it does not read as ad-hoc inline detail.

Dimension	Reasoning	Score
Conciseness	The content assumes Claude's intelligence (no explanation of HF/Megatron/TP/PP basics), but every line is duplicated bilingually in English and Chinese, roughly doubling token cost for a single-language context — a clear tightening opportunity.	2 / 3
Actionability	Provides copy-paste-ready commands, exact file paths, environment-variable tables with defaults, named functions, weight-mapping tables, and concrete thresholds (e.g., cosine > 0.9999).	3 / 3
Workflow Clarity	Multi-step processes are clearly sequenced with explicit validation checkpoints (path existence checks, conversion-success verification) and feedback loops (priority-ordered diagnosis, 'fix weight_consistency first' troubleshooting).	3 / 3
Progressive Disclosure	No bundle files exist and the skill is a ~370-line monolith covering test structure, fixtures, six test specs, env vars, run instructions, troubleshooting, and weight mapping — well-organized but content that could be split into separate references is inline.	2 / 3
	Total	10 / 12 Passed

Description

72%

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 clear niche with good natural trigger terms and low conflict risk, but it lacks an explicit 'Use when...' trigger clause, capping completeness at 2.

Suggestions

Add an explicit 'Use when...' clause stating when to invoke the skill (e.g., 'Use when validating HF vs Megatron checkpoint consistency or changing TP/PP layout without retraining').

Enumerate a few more concrete actions in the description (e.g., 'runs weight, vision-encoder, and end-to-end LLM consistency tests') to lift specificity to 3.

Dimension	Reasoning	Score
Specificity	Names the domain ('LLaVA-OneVision2 HF vs Megatron consistency checks') and some concrete actions ('running and interpreting', 'across TP and PP settings'), but does not list multiple specific concrete operations like a fuller enumeration would.	2 / 3
Completeness	States clearly what the skill does, but there is no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 2 per the judging guidelines.	2 / 3
Trigger Term Quality	For this repo's audience, 'LLaVA-OneVision2', 'HF vs Megatron', 'consistency checks', and 'TP and PP' are exactly the natural terms a user would say when they need this skill.	3 / 3
Distinctiveness Conflict Risk	The 'LLaVA-OneVision2 HF vs Megatron consistency checks across TP and PP settings' niche is highly specific and unlikely to trigger for the wrong skill.	3 / 3
	Total	10 / 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/llava-onevision2-consistency/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.