CtrlK
BlogDocsLog inGet started
Tessl Logo

verify-conversion

E2E gate — verifies that applied patches produce a working, numerically sane end-to-end inference through the OpenVINO plugin. Handles HuggingFace/optimum-intel, native OV conversion (ovc/convert_model), and ONNX. Used by orchestrators as the mandatory gate before any PR is published.

64

Quality

76%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./.github/agents-prototype/skills/verify-conversion/SKILL.md
SKILL.md
Quality
Evals
Security

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.

A tight, code-driven verification gate with executable examples for each conversion path, explicit numerical-sanity validation, and clear failure handling. The main dings are minor: a few placeholders in code, a Step 1→Step 3 numbering gap, and a slightly padded intro blockquote.

Suggestions

Fix the step numbering (label Path A/B as "Step 2 — Convert and run inference") so the sequence reads Step 1 → Step 2 → Step 3 without a gap.

Tighten the intro blockquote to the hard rule plus the one-line rationale, dropping the parenthetical plugin-issue enumeration.

Replace `# torch_model = ...` and "adjust shape" with a concrete, runnable toy-model example (or an explicit note that the placeholder is context-dependent) to push actionability toward fully copy-paste-ready.

DimensionReasoningScore

Conciseness

Code-first and mostly lean with a decision table and minimal prose, matching the "efficient; minor instances of over-explanation that could be trimmed" anchor; not a 5 because the blockquote intro (e.g., "plugin-level issues (wrong kernel output, silent data corruption...) are only caught by an actual inference run") is slightly padded.

4 / 5

Actionability

Provides concrete, mostly copy-paste-ready code for every path (AutoConfig detection, optimum-cli export, OVModelForCausalLM inference, ONNX/PyTorch/ovc compile+infer, sanity-check functions, JSON schema), matching the "mostly executable; minor gaps" anchor; not a 5 because of placeholders like `# torch_model = ...`, "adjust shape", and unset `$MODEL_ID`/`$TASK` env vars.

4 / 5

Workflow Clarity

Clear sequence (determine path → convert+infer → sanity check → report) with an explicit validation checkpoint (Step 3 numerical sanity) and feedback loops (retry int4 on OOM, "do not silently swallow the failure"), matching the "clear sequence with most checkpoints" anchor; not a 5 because numbering jumps from Step 1 to Step 3 with no labeled Step 2, and the final gate decision is stated in Result reporting rather than as an explicit checklist.

4 / 5

Progressive Disclosure

Well-organized into clearly headed sections (Step 1, Path A, Path B, Step 3, Result reporting), self-contained with no nested references and easy header-based navigation, matching the "good structure; most content appropriately placed" anchor; not a 5 because it is a single monolithic ~210-line file where per-path code blocks could optionally be split into reference scripts.

4 / 5

Total

16

/

20

Passed

Description

78%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 specific, well-scoped description that clearly communicates the skill's niche (OpenVINO conversion E2E verification gate) and when it applies. It would reach the top band with an explicit "Use when..." trigger clause and slightly broader synonym coverage.

Suggestions

Add an explicit "Use when..." trigger phrase (e.g., "Use when verifying that a model conversion to OpenVINO produces working, numerically sane inference before publishing a PR") to lift completeness.

Include a few natural synonyms/file extensions (e.g., "OpenVINO IR", ".xml/.bin", "model conversion") to broaden trigger-term coverage.

Mention the result artifact (writes verify_result.json) so the capability list is comprehensive.

DimensionReasoningScore

Specificity

Names the domain and several concrete actions — "verifies that applied patches produce a working, numerically sane end-to-end inference", "Handles HuggingFace/optimum-intel, native OV conversion (ovc/convert_model), and ONNX", "mandatory gate before any PR is published" — matching the "lists several specific actions; minor gaps" anchor; not a 5 because coverage omits the result-reporting action.

4 / 5

Completeness

Clearly states what (verify conversion + numerically sane E2E inference) and when ("Used by orchestrators as the mandatory gate before any PR is published"), matching the "both what and when; when could be more explicit" anchor; not a 5 because it lacks a concrete "Use when..." trigger phrase and frames the when around orchestrators rather than user triggers.

4 / 5

Trigger Term Quality

Good domain keyword coverage with natural-for-the-niche terms ("OpenVINO plugin", "optimum-intel", "ovc/convert_model", "ONNX", "E2E gate", "inference", "conversion"); not a 5 because synonyms like "OpenVINO IR" and generic "model conversion" phrasings are missing.

4 / 5

Distinctiveness Conflict Risk

Clear niche — OpenVINO conversion E2E verification gate — with distinct triggers ("E2E gate", "numerically sane end-to-end inference through the OpenVINO plugin") and minimal conflict risk; not lower because it is sharply distinguishable from general conversion skills.

5 / 5

Total

17

/

20

Passed

Validation

100%

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

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
openvinotoolkit/openvino
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.