modal-serverless-gpu

Run governed Modal sandbox jobs through OpenScience's compute_job JobBroker. Use for one-off CPU/GPU commands, explicit file uploads and captures, dependency provisioning, resource selection, approval, dispatch, and results. This skill does not install or invoke the Modal Python SDK or CLI.

Quality

86%

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

86%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 concise, highly actionable, and well-sequenced with an approval checkpoint, making it strong on token efficiency and execution guidance. Its main weakness is progressive disclosure: real reference files are present but never linked or navigated from the body, leaving them effectively undiscoverable.

Suggestions

Add a short '## Further reference' section with explicit links to references/advanced-patterns.md, references/examples-catalog.md, and references/troubleshooting.md, noting which to consult for which situation.

Resolve the contradiction between 'legacy reference files... are not execution instructions' and the existence of clearly current references (e.g. examples-catalog.md, troubleshooting.md); either relabel them or clarify which are legacy vs. current.

Add an explicit validate/retry feedback loop for job failures (e.g., on non-zero exit, inspect the returned logs, fix inputs/packages, and re-dispatch) so the workflow reaches a score-5 checkpoint pattern.

Dimension	Reasoning	Score
Conciseness	The body is lean and assumes competence: it gives rules, a GPU selection table, and a complete example without explaining what Modal or sandboxes are. Every section earns its place.	5 / 5
Actionability	Provides a fully executable JSON compute_job call, concrete shell commands ('python analysis.py'), explicit package pins ('numpy==2.3.2'), and a concrete GPU table—copy-paste ready for the common case.	5 / 5
Workflow Clarity	The 'Preparing a job' section is a clear numbered sequence with an explicit approval checkpoint and status-reporting step, but it lacks an explicit validate-on-error feedback loop for runtime failures, so it sits just below the score-5 anchor.	4 / 5
Progressive Disclosure	Bundle files exist in references/ (advanced-patterns.md, examples-catalog.md, troubleshooting.md) but the body never links or signals them by name; the only mention is a vague 'legacy reference files in this skill directory' that actively disclaims them, so navigation is unclear and content split is not properly surfaced.	3 / 5
	Total	17 / 20 Passed

Description

87%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.

The description is specific, complete, and clearly differentiated, with an explicit 'Use for...' trigger and a useful negative-boundary statement. It is slightly jargon-heavy ('JobBroker') and could swap in more user-natural synonyms, but it lands at the top of the scale.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions ('explicit file uploads and captures, dependency provisioning, resource selection, approval, dispatch, and results') rather than vague language, though they are compact and slightly less granular than the score-5 anchor.	4 / 5
Completeness	Clearly answers 'what' (runs governed Modal sandbox jobs via compute_job) and 'when' with an explicit 'Use for one-off CPU/GPU commands...' trigger clause, plus a boundary statement.	5 / 5
Trigger Term Quality	Includes natural terms like 'one-off CPU/GPU commands', 'file uploads and captures', and 'Modal', but leans on jargon ('JobBroker') and misses some common synonyms a user would naturally say; not quite comprehensive.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (Modal through OpenScience Compute) and explicitly distinguishes itself from the standalone Modal SDK/CLI, minimizing wrong-skill triggering.	5 / 5
	Total	18 / 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/cloud-compute/modal/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.