Tessl • The package manager for agent skills

SkillPublic

optimize-for-gpu

GPU-accelerates scientific Python on NVIDIA hardware and verifies that the result is correct and faster. Use for CUDA/GPU optimization; CPU-bound NumPy, SciPy, pandas, scikit-learn, NetworkX, scikit-image, vector-search, image-processing, graph, simulation, or file-I/O workloads; CuPy, cuDF, cuML, cuGraph, cuVS, cuCIM, KvikIO, Warp, Newton, Numba-CUDA, or RAFT questions; and profiling, memory-transfer, kernel, or multi-GPU bottlenecks. Also use when large data-parallel Python code is slow and GPU acceleration is a plausible option, even if the user does not name CUDA.

Quality

91%

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

The canonical home for this skill is optimize-for-gpu in K-Dense-AI/scientific-agent-skills

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.

An excellent overview body: a disciplined evidence-driven workflow with validation gates, concrete tooling, and clean one-level-deep progressive disclosure into an exhaustive reference set. The one notable weakness is the padded 'When This Skill Applies' trigger list, which is largely redundant with the description and itself.

Suggestions

Consolidate the 22 'When This Skill Applies' bullets into ~8-10 grouped triggers (e.g., merge the three graph-analytics bullets and the three simulation/robotics bullets) since the section largely duplicates the frontmatter description.

Remove near-duplicate triggers (image processing appears in both the scikit-image and whole-slide imaging bullets; file I/O in both the fromfile and S3/HTTP bullets) and keep one canonical phrasing per domain.

Move the exhaustive legacy-status details for cuxfilter and cuSpatial fully into their reference files, keeping only a one-line status marker in the main tables.

Dimension	Reasoning	Score
Conciseness	The workflow, decision tables, and Important Notes are lean and imperative with no concept re-teaching, but the 22-bullet 'When This Skill Applies' section is substantially redundant (graph analytics appears in ~3 bullets, simulation in ~3, image processing in 2-3) and duplicates the frontmatter description, so the body could be noticeably tightened.	3 / 5
Actionability	The body provides a copy-paste-runnable cupyx.profiler.benchmark snippet, a concrete workload-to-library decision table, named tools and modes (nsys, ncu, %gpu_timeit, cudf.pandas, cuml.accel), and specific per-step instructions covering the common cases.	5 / 5
Workflow Clarity	The seven-step workflow is clearly sequenced with explicit validation checkpoints ('Validate semantics before speed', CPU/GPU output comparison, recall@k for ANN indexes) and a keep/revise/reject feedback loop gated on correctness checks.	5 / 5
Progressive Disclosure	The body is a well-organized overview with 15 clearly signaled one-level-deep reference files, each mapped via a 'When to Read' table, and every cited reference path was verified to exist in references/.	5 / 5
	Total	18 / 20 Passed

Description

95%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 explicitly states both what the skill does and when to use it, with comprehensive natural trigger terms and a distinct GPU/NVIDIA niche. The only weakness is a narrow set of distinct actions (accelerate + verify), which keeps specificity just below the top anchor.

Dimension	Reasoning	Score
Specificity	Concrete actions are stated ('GPU-accelerates scientific Python on NVIDIA hardware and verifies that the result is correct and faster') with comprehensive domain and library coverage, but the distinct action list is narrow (accelerate plus verify), leaving minor gaps versus the comprehensive multi-action anchor.	4 / 5
Completeness	The first sentence explicitly answers 'what' (GPU-accelerates scientific Python and verifies correctness and speed) and 'Use for... Also use when...' explicitly answers 'when' with concrete trigger phrases.	5 / 5
Trigger Term Quality	Natural user phrasing is comprehensively covered — 'NumPy, SciPy, pandas, scikit-learn', 'CUDA/GPU optimization', 'profiling, memory-transfer, kernel, or multi-GPU bottlenecks' — plus library names and a synonym-style catch-all ('even if the user does not name CUDA').	5 / 5
Distinctiveness Conflict Risk	A clear NVIDIA/CUDA optimization niche with distinct triggers; the broad catch-all clause is qualified by 'GPU acceleration is a plausible option', keeping overlap with generic optimization skills minimal.	5 / 5
	Total	19 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/coding/optimize-for-gpu/SKILL.md
Revision: 6a85742

Reviewed: 4 days ago

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.

optimize-for-gpu

Table of Contents

Is this your skill?