llava-onevision2-consistency

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

Quality

72%

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

75%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 body is highly actionable with concrete commands, env vars, and clear run/diagnosis workflows, and is well-structured for navigation. Its main weakness is conciseness: systematic bilingual duplication doubles token load, and some detailed reference material could be split into separate files.

Suggestions

Reduce bilingual token cost by, e.g., interleaving EN/中文 inline per bullet rather than duplicating whole sections, or moving one language to a reference file, to improve conciseness.

Split the long per-test detail, weight-mapping table, and failure-cause table into reference files (e.g., TESTS.md, WEIGHT_MAP.md) with one-level-deep links from SKILL.md to improve progressive disclosure.

Add an explicit validate-then-proceed checkpoint (e.g., 'Confirm test_weight_consistency passes before trusting other tests') as a numbered workflow step rather than only a troubleshooting note.

Dimension	Reasoning	Score
Conciseness	Per-language content is lean and concrete, but the systematic bilingual duplication of every section roughly doubles tokens; while intentional for a 'Bilingual guide', it is noticeable verbosity that could be tightened, sitting between the 2 and 3 anchors.	3 / 5
Actionability	Fully executable, copy-paste-ready commands throughout (run_consistency_tests.sh invocations, pytest commands, conversion scripts with args, env-var table with defaults, exact thresholds like cosine > 0.9999) covering the common cases.	5 / 5
Workflow Clarity	Clear run workflows for multiple scenarios plus a numbered 5-step 'what the script does' sequence and a 7-step troubleshooting checklist; most checkpoints present, with minor validation gaps and no destructive-operation cap applying.	4 / 5
Progressive Disclosure	No bundle files exist and the single SKILL.md is well-organized with clear headers, but at ~371 lines some detail (per-test breakdowns, weight-mapping and failure-cause tables) could be split into reference files, leaving minor organization gaps.	4 / 5
	Total	16 / 20 Passed

Description

70%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 specific and distinctive, naming concrete actions within a well-defined niche, but it omits an explicit 'Use when' trigger clause in the frontmatter, which caps its completeness. Trigger-term coverage is good but could add common synonyms.

Suggestions

Add an explicit 'Use when...' trigger clause to the frontmatter description (e.g., 'Use when validating checkpoint consistency or converting between HF and Megatron formats') to raise completeness.

Include natural synonyms users say, such as 'tensor parallel', 'pipeline parallel', 'checkpoint conversion', or 'model conversion', to broaden trigger-term coverage.

Spell out the concrete validation actions (e.g., compare weights, compare logits, run pytest suite) to push specificity toward a comprehensive action list.

Dimension	Reasoning	Score
Specificity	Names the domain (LLaVA-OneVision2 HF vs Megatron) plus concrete actions ('running and interpreting consistency checks across TP and PP settings'), giving several specific actions with minor coverage gaps rather than a comprehensive list.	4 / 5
Completeness	Provides a clear 'what' but the frontmatter description itself contains no 'Use when...' trigger clause; per the rubric a missing explicit trigger caps completeness at 3 even though the body adds the trigger.	3 / 5
Trigger Term Quality	Includes natural domain terms a validation engineer would say ('consistency checks', 'HF vs Megatron', 'TP and PP settings') with good coverage, though missing common synonyms like 'tensor/pipeline parallel' or 'checkpoint conversion'.	4 / 5
Distinctiveness Conflict Risk	Targets a clear niche (LLaVA-OneVision2 HF/Megatron consistency across TP/PP) with distinct triggers and minimal overlap risk with other skills.	5 / 5
	Total	16 / 20 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: f1b6702

Reviewed: 2 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.