skypilot-multi-cloud-orchestration

Multi-cloud orchestration for ML workloads with automatic cost optimization. Use when you need to run training or batch jobs across multiple clouds, leverage spot instances with auto-recovery, or optimize GPU costs across providers.

Quality

80%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

72%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 highly actionable, well-structured reference with real, cleanly organized bundle files. The main weakness is conciseness padding and the absence of validation checkpoints around spot/batch operations, which caps workflow clarity.

Suggestions

Add explicit verification steps for batch and destructive operations — e.g. after `sky jobs launch` note checking `sky jobs queue`/`sky jobs logs` for failure before proceeding, and after `sky down` confirm `sky status` shows termination.

Trim the 'When to use SkyPilot' and 'Key features' bullet lists, which restate capabilities already implied by the description and code examples; keep only what Claude cannot infer.

Consolidate the enumerated GPU accelerator list (T4/L4/A10G/L40S/A100/H100) into a pointer or a compact table rather than a standalone block, since model availability is cloud/version-specific and time-sensitive.

Dimension	Reasoning	Score
Conciseness	Mostly efficient code-forward structure, but it pads with lists of features and concepts (e.g. the 'When to use SkyPilot' bullets, 'Key features' with savings figures, GPU model enumerations) that largely restate what Claude can infer; some tightening would help.	3 / 5
Actionability	Copy-paste-ready YAML and shell commands throughout — install, hello world, GPU fallbacks, distributed torchrun, managed jobs, file mounts, Sky Serve — covering the common cases with specific executable examples.	5 / 5
Workflow Clarity	Sections are well sequenced, but batch/destructive operations (spot jobs, cluster launch/down, multi-job sweeps) lack explicit validation or verification checkpoints and the cap for missing feedback loops applies.	3 / 5
Progressive Disclosure	Body is a clear overview with two well-signaled one-level-deep references — [Advanced Usage](references/advanced-usage.md) and [Troubleshooting](references/troubleshooting.md) — both of which are real files, with content appropriately split.	5 / 5
	Total	16 / 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: third-person, concrete, with an explicit 'Use when...' trigger and comprehensive action coverage. Minor gains available from adding common synonyms and provider-specific terms to the trigger list.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'run training or batch jobs across multiple clouds, leverage spot instances with auto-recovery, or optimize GPU costs across providers' — with comprehensive coverage of the skill's capabilities.	5 / 5
Completeness	Clearly answers both 'what' ('Multi-cloud orchestration for ML workloads with automatic cost optimization') and 'when' via an explicit 'Use when you need to...' clause with concrete trigger phrases.	5 / 5
Trigger Term Quality	Good natural keywords ('training or batch jobs', 'spot instances', 'GPU costs', 'multiple clouds') a user would actually say, but missing common synonyms or explicit provider/region terms that would round out coverage.	4 / 5
Distinctiveness Conflict Risk	The SkyPilot/multi-cloud-cost-optimization niche is fairly distinct, but the broad 'ML workloads across multiple clouds' framing could overlap with general cloud-orchestration or Ray skills.	4 / 5
	Total	18 / 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
skill_md_line_count	SKILL.md is long (511 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/cloud-compute/skypilot/SKILL.md
Commit: 3a6c3a9

Reviewed: 1 day 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.