modal

Serverless GPU cloud for ML jobs and model APIs.

Quality

61%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

High

Do not use without reviewing

Fix and improve this skill with Tessl

tessl review fix ./optional-skills/mlops/modal/SKILL.md

Quality

Content

68%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 content is a lean, highly actionable quick reference dominated by executable code and tables, correctly deferring advanced usage and troubleshooting to real bundle files. Its main weaknesses are a slightly overloaded body and missing validation/verification steps around batch and deployment workflows.

Suggestions

Add verification steps to the quick start and parallel-processing workflows (e.g. confirm `modal token` / run `.local()` before fanning out `.map()`, and check results for failures before proceeding).

Move lower-frequency sections (secrets, scheduling, GPU table detail) into references/advanced-usage.md to slim the SKILL.md body toward a true overview.

Replace undefined placeholders (download_model(), model.predict) in the volume and endpoint examples with runnable code so every snippet is copy-paste executable.

Dimension	Reasoning	Score
Conciseness	The body is dense with tables and executable code with no explanations of concepts Claude already knows, but sections like "Key features" ("Sub-second cold starts: Rust-based infrastructure") and "Use alternatives instead" carry marketing-style padding that could be trimmed or moved to references.	4 / 5
Actionability	Nearly all snippets are copy-paste ready (installation, hello-world, GPU spec patterns, image builds, secret creation), but a few examples reference undefined placeholders like download_model(), load_from_path(), and model.predict(text), keeping them from fully executable.	4 / 5
Workflow Clarity	The quick start is well sequenced (install → hello world → endpoint → execution modes) with a debugging section, but the parallel .map() fan-out over 1000 items is a batch operation with no validation or verification steps, which caps workflow clarity at 3 per the batch-operations rule.	3 / 5
Progressive Disclosure	Two real one-level-deep reference files (advanced-usage.md, troubleshooting.md) are clearly signaled in a References section, but the ~340-line body keeps material (secrets, scheduling, GPU detail) inline that could live in references, leaving minor organization gaps.	4 / 5
	Total	15 / 20 Passed

Description

53%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 concise and names a distinct niche, but it is a noun-phrase label rather than a capability-plus-trigger statement. It lacks any "Use when..." guidance and misses key natural trigger terms like "Modal" itself, "training", and "inference".

Suggestions

Add an explicit trigger clause, e.g. "Use when the user mentions Modal, serverless GPUs, running training/inference jobs in the cloud, or deploying models as APIs."

Convert the label into concrete actions, e.g. "Run ML training and inference jobs on serverless GPUs and deploy models as auto-scaling APIs."

Include natural synonyms and brand terms (Modal, GPU cloud, H100/A100, model deployment) that users would actually say when they need this skill.

Dimension	Reasoning	Score
Specificity	"Serverless GPU cloud for ML jobs and model APIs" names the domain plus two capability areas but uses no action verbs and omits training, inference, and deployment specifics, matching the 'names domain and 1-2 concrete actions' anchor.	3 / 5
Completeness	The 'what' is clear (serverless GPU cloud for ML jobs and model APIs) but there is no "Use when..." clause or equivalent trigger guidance, so completeness is capped at 3 per the judging guideline.	3 / 5
Trigger Term Quality	Natural terms like "serverless", "GPU", "ML jobs", and "model APIs" are present, but common variations users would actually say ("Modal", "training", "inference", "deploy", specific GPU names) are missing, matching the 'some relevant keywords but missing variations' anchor.	3 / 5
Distinctiveness Conflict Risk	The serverless-GPU-for-ML niche is mostly distinct with minor overlap risk against generic cloud-deployment or model-serving skills, matching the 'mostly distinct; minor overlap risk' anchor.	4 / 5
	Total	13 / 20 Passed

Validation

81%

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 — 13 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	13 / 16 Passed

Repository: NousResearch/hermes-agent
Path: optional-skills/mlops/modal/SKILL.md
Commit: 68fa7e9

Reviewed: 3 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.