litert-model-equivalence-test

Validates equivalence between LiteRT models (litert_lm) and PyTorch models (transformers). Use when you need to verify that an exported LiteRT model produces the same outputs as the original Hugging Face model. Supports multi-turn conversations and custom prompts.

Quality

85%

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

Quality

Content

82%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.

A lean, highly actionable body with copy-paste commands and good section structure. It could be tightened by factoring the shared bazel target out of the examples and adding brief guidance on interpreting equivalence-test results.

Suggestions

Factor the repeated bazel target path into a variable or state it once, then show only the differing flags in each example to reduce token overhead.

Add a short section on interpreting results (what a pass vs. failure looks like, and how to react when equivalence fails) to strengthen the validation workflow.

Trim the opening sentence ('This skill provides instructions for running equivalence tests...') since it restates the frontmatter description.

Dimension	Reasoning	Score
Conciseness	Efficient and free of concept over-explanation, but the full bazel target path is repeated verbatim in every example and the opening sentence restates the frontmatter description.	4 / 5
Actionability	Fully executable, copy-paste-ready bazel commands with a complete flag reference and examples covering single-turn, multi-turn, prompt-file, externalized-embedder, and NPU variants.	5 / 5
Workflow Clarity	The single bazel-run action is unambiguous, but for a validation skill there is no guidance on interpreting pass/fail results or what to do when equivalence fails.	4 / 5
Progressive Disclosure	Well-organized into Usage, Flags, and Examples sections with no nested references and no bundle files, though the body exceeds 50 lines and repeats the command rather than factoring it out.	4 / 5
	Total	17 / 20 Passed

Description

87%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.

A strong description that clearly states both its purpose and an explicit trigger clause, with concrete capabilities and a distinct niche. Minor improvement could come from adding natural synonyms like 'compare' or 'match outputs'.

Dimension	Reasoning	Score
Specificity	Names the domain and several concrete actions ('Validates equivalence', 'verify that an exported LiteRT model produces the same outputs', 'Supports multi-turn conversations and custom prompts'), with only minor coverage gaps.	4 / 5
Completeness	Explicitly answers both what ('Validates equivalence between LiteRT models... and PyTorch models') and when ('Use when you need to verify that an exported LiteRT model produces the same outputs...').	5 / 5
Trigger Term Quality	Good natural keyword coverage ('LiteRT models', 'PyTorch models', 'transformers', 'Hugging Face model', 'multi-turn conversations'), though common synonyms like 'compare' or 'match outputs' are absent.	4 / 5
Distinctiveness Conflict Risk	A clear niche (LiteRT vs PyTorch/Hugging Face equivalence) with distinct triggers and minimal overlap risk with other skills.	5 / 5
	Total	18 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: google-ai-edge/litert-torch
Path: litert_torch/generative/export_hf/experimental/validation/SKILL.md
Commit: 70d7619

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.