Name: warp-eval
Rating: 68 (1 reviews)
Author: NVIDIA

warp-eval

Evaluate whether an existing hot path is a credible NVIDIA Warp candidate. Use for irregular or spatial queries, particle or geometry simulation, branch-heavy loops, many small launches, host fallbacks, or large intermediates. CPU-only code and absent GPU dependencies are normal unless NVIDIA is prohibited. Exclude required cross-vendor or CPU-only deployment, vendor-lowered dense or NN layers, general Warp API questions, and already-selected Warp kernels. Contribution policy alone is not exclusion.

Quality

83%

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

85%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 tightly structured, high-stakes evaluation workflow with nine sequenced stages, explicit gates, validation feedback loops, and a clean one-level-deep reference hierarchy. It is only slightly verbose from intentional reinforcement of authorization/gate rules, and a few procedural steps lack executable commands.

Suggestions

Consolidate the gate definitions and authorization rules so each is stated once in its canonical location with cross-references, reducing repetition between the Hard Rules and per-stage instructions.

Add one or two concrete example commands for the profiling and baseline-measurement steps (e.g., a representative profiler invocation pattern) to lift actionability from mostly-executable to copy-paste ready.

Tighten the recurring authorization phrasing in stages 2, 5, 6, and 7 to a single shared checkpoint reference to trim tokens without losing the gate.

Dimension	Reasoning	Score
Conciseness	The body is dense and assumes Claude's competence — it never explains basic GPU/profiling concepts — but authorization and gate definitions recur across the Hard Rules and individual stages, so it is efficient rather than maximally lean, sitting below the level-5 "every token earns its place" bar.	4 / 5
Actionability	Concrete, copy-paste commands are present ("uv run python scripts/validate_report_schema.py <report-directory>", a run_script example, a scripts table with arguments, exact gate criteria), but several steps ("Profile with the project's own profiler", prototyping) are procedural rather than executable code, leaving minor gaps versus the level-5 anchor.	4 / 5
Workflow Clarity	Nine numbered stages with explicit stop/abort/gate conditions, a gate table, an output-state table, and a validate-fix-revalidate feedback loop ("Run the validator once the first measurement establishes a census/table, and again before delivery. Fix every error." plus "deliberately break a branch to prove the tests fail") match the level-5 anchor for sequenced process with validation checkpoints and feedback loops.	5 / 5
Progressive Disclosure	The body is an overview that points to seven one-level-deep reference files, three scripts, and one asset template via clearly signaled markdown links, all of which exist in the bundle; content is appropriately split with easy navigation, matching the level-5 anchor.	5 / 5
	Total	18 / 20 Passed

Description

82%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 precise, third-person description that cleanly states one concrete evaluation action and pairs it with explicit "Use for/Exclude" trigger guidance. Its only limitation is that, as a single-purpose screening skill, it cannot list multiple distinct actions, capping the specificity dimension.

Dimension	Reasoning	Score
Specificity	The description names the domain ("NVIDIA Warp candidate") and one concrete action — "Evaluate whether an existing hot path is a credible NVIDIA Warp candidate" — framed precisely with include/exclude criteria, but it is a single-action skill rather than the "several specific actions" the level-4 anchor requires.	3 / 5
Completeness	It explicitly answers both what ("Evaluate whether an existing hot path is a credible NVIDIA Warp candidate") and when via a concrete "Use for ... Exclude ..." trigger clause with concrete trigger phrases, matching the level-5 anchor.	5 / 5
Trigger Term Quality	Natural domain terms a performance-minded user would say appear throughout ("NVIDIA Warp", "particle or geometry simulation", "branch-heavy loops", "many small launches", "host fallbacks", "large intermediates", "GPU"); falls short of level 5 only because it lacks broader synonym/extension coverage (e.g. CUDA, kernel, .py).	4 / 5
Distinctiveness Conflict Risk	It carves a clear niche (Warp candidacy screening) with precise include/exclude boundaries ("Exclude required cross-vendor or CPU-only deployment, vendor-lowered dense or NN layers, general Warp API questions, and already-selected Warp kernels"), giving minimal overlap with other skills.	5 / 5
	Total	17 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning

	Total	14 / 16 Passed

Repository: NVIDIA/warp
Path: skills/warp-eval/SKILL.md
Commit: 238de60

Reviewed: 5 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.