CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

veomni-new-op

Add or optimize a tensor-level kernel in veomni/ops, register its variants, integrate it with models, and add numerical and registry tests. Trigger: 'add op', 'new kernel', 'add attention variant', 'new fused op', 'add triton kernel', 'optimize operator'.

Quality

90%

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

SKILL.md
Quality
Evals
Security

Quality

Content

90%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 an exemplary lean, repo-specific workflow: phased design-to-finalize structure, executable registration code, concrete test commands, and explicit pitfalls. Weakest points are the absence of an explicit error-recovery loop after validation failures and inline policy detail that could live in a reference file.

DimensionReasoningScore

Conciseness

The ~165-line body is dense and entirely repo-specific (registry identity tuple, patchgen rules, NPU eager restrictions, cache rules) with zero explanation of concepts Claude already knows; every section carries project constraints, matching the lean/efficient anchor.

5 / 5

Actionability

It provides a copy-paste-ready register_op example covering both the wrapper and raw-pair forms, an exact test command (pytest -q tests/ops/<kernel_name>/ tests/ops/base/), and concrete verification commands (OP_REGISTRY.list_registered(...)), with the forward/backward contract precisely specified; specific examples cover the common cases.

5 / 5

Workflow Clarity

Five clearly sequenced phases plus a 'Before You Start' section and validation checkpoints (per-input backward parity tests, /veomni-review, make quality, registry verification) give a clear sequence; however, no explicit fix-and-rerun feedback loop after validation failures keeps it below the anchor-5 example.

4 / 5

Progressive Disclosure

Sections are well-organized with clearly signaled one-level-deep references to repo docs (veomni/ops/README.md, docs/design/op_selection.md, docs/design/module_config.md, the new-model coverage policy link) and no bundle files exist; dense policy sections like 'Configuration ownership' are inlined where a reference-file split would help, fitting the minor-organization-gaps anchor.

4 / 5

Total

18

/

20

Passed

Description

88%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: it states the full workflow concretely in third-person imperative voice and provides an explicit trigger clause with natural, domain-specific phrases. The only improvements are broader trigger synonyms and slightly less generic trigger wording to reduce overlap risk.

DimensionReasoningScore

Specificity

"Add or optimize a tensor-level kernel in veomni/ops, register its variants, integrate it with models, and add numerical and registry tests" lists four concrete actions covering the full lifecycle (implement, register, integrate, test), matching the comprehensive-coverage anchor with no meaningful gaps for this scope.

5 / 5

Completeness

The description explicitly answers both 'what' (the four concrete actions) and 'when' via a dedicated "Trigger: 'add op', 'new kernel', ..." clause with concrete trigger phrases, structurally matching the anchor-5 example.

5 / 5

Trigger Term Quality

Six natural trigger phrases including domain-specific variants ('add triton kernel', 'new fused op', 'add attention variant') give good keyword coverage, but common synonyms such as 'custom op', 'register op', or 'speed up op' are missing, so it falls short of the comprehensive-synonym anchor.

4 / 5

Distinctiveness Conflict Risk

The niche (tensor-level kernels in veomni/ops) is clearly distinct, but generic triggers like 'new kernel' and 'optimize operator' create minor overlap risk with closely related repo skills, fitting the 'mostly distinct' anchor rather than the minimal-conflict one.

4 / 5

Total

18

/

20

Passed

Validation

93%

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

CriteriaDescriptionResult

relative_links

Relative link issues: 1 suspicious

Warning

Total

15

/

16

Passed

Repository

ByteDance-Seed/VeOmni

Path

.agents/skills/veomni-new-op/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.