CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

merge-ov2

Bilingual guide for merging ViT + LLM into LlavaOnevision2 HF checkpoint and validating weight/inference consistency

Quality

73%

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/merge-ov2/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

81%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 executable commands and code, a clear validated workflow, and strong error-recovery guidance; its main weakness is token efficiency from bilingual duplication and the absence of any bundle-file offloading despite the length.

DimensionReasoningScore

Conciseness

Per language the prose is lean and assumes Claude's competence (no basic-concept padding), but every section is duplicated in English and Chinese, roughly doubling the token cost; this fits 'mostly efficient but includes some unnecessary explanation or could be tightened'.

3 / 5

Actionability

Bash commands are complete with docker exec, PYTHONPATH, and real paths, and the Python validation scripts are copy-paste ready with concrete asserts and thresholds, matching 'fully executable; copy-paste ready code'.

5 / 5

Workflow Clarity

The merge → validate → Megatron-conversion sequence is explicit, the validate subcommand and three manual test scripts provide validation checkpoints with thresholds, and Known Issues supplies error-recovery feedback loops (OOM fallback, modeling drift).

5 / 5

Progressive Disclosure

Sections are well-organized with clear headers and clearly signaled cross-references (the llava-onevision2-consistency skill, merge_ov2/utils.py), but ~680 lines are inlined in SKILL.md with no bundle files, so content like the manual scripts and known-issues could be split out, keeping it just below 5.

4 / 5

Total

17

/

20

Passed

Description

57%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 states a clear, specific purpose and occupies a distinct niche, but it omits any 'Use when...' trigger guidance and only offers two concrete actions, capping completeness and specificity at 3.

Suggestions

Add an explicit 'Use when...' clause naming the natural trigger phrases (e.g. 'Use when merging a standalone ViT encoder and LLM into a LlavaOnevision2 checkpoint, or validating merge parity').

List a couple more concrete actions (e.g. QKV fusion, weight-key remapping, dry-run coverage) to lift specificity toward 4-5.

Include common natural-language synonyms and file artifacts users might reference (e.g. 'safetensors checkpoint', 'LlavaOnevision2ForConditionalGeneration') to broaden trigger-term coverage.

DimensionReasoningScore

Specificity

Names the domain ('merging ViT + LLM into LlavaOnevision2 HF checkpoint') and two concrete actions ('merging' and 'validating weight/inference consistency'), matching the anchor for 1-2 concrete actions rather than the several-action coverage of 4.

3 / 5

Completeness

It clearly answers 'what' but contains no 'Use when...' clause or equivalent explicit trigger guidance; per the rubric a missing trigger clause caps completeness at 3.

3 / 5

Trigger Term Quality

Relevant keywords appear ('merging', 'checkpoint', 'validating', 'consistency', 'ViT', 'LLM') but natural user phrasing and synonyms are missing, fitting 'some relevant keywords but missing common variations' rather than the comprehensive coverage of 4-5.

3 / 5

Distinctiveness Conflict Risk

The niche is highly specific (LlavaOnevision2 HF checkpoint merge + weight/inference validation) with distinct triggers and minimal overlap with other skills.

5 / 5

Total

14

/

20

Passed

Validation

81%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 13 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (684 lines); consider splitting into references/ and linking

Warning

metadata_version

'metadata.version' is missing

Warning

relative_links

Relative link issues: 1 missing

Warning

Total

13

/

16

Passed

Repository

EvolvingLMMs-Lab/LLaVA-OneVision-2

Path

.opencode/skills/merge-ov2/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.