CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

serverless-modal

Run GPU workloads on Modal — training, fine-tuning, inference, batch processing. Zero-config serverless: no SSH, no Docker, auto scale-to-zero. Use when user says "modal run", "modal training", "modal inference", "deploy to modal", "need a GPU", "run on modal", "serverless GPU", or needs remote GPU compute.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

63%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-sequenced operational guide whose main costs are bulk and redundancy: everything lives inline in one large file, key warnings and commands are duplicated, and time-sensitive pricing is embedded without staleness handling. Splitting reference material into bundle files and consolidating duplicated sections would raise both conciseness and progressive disclosure.

Suggestions

Deduplicate content: merge the two near-verbatim card-binding/spending-limit warnings (Authentication blockquote and closing 'Cost protection' blockquote) into one, and consolidate cost-estimation guidance from '!! Cost Estimation Required !!' and Step 1 into a single section.

Offload bulk material into bundle reference files (e.g., references/pricing.md for the GPU price and benchmark tables, references/patterns.md for Patterns C–F), keeping SKILL.md as a lean overview with clearly signaled one-level-deep links.

Handle time-sensitive data: replace exact $/sec prices with rounded $/hr guidance plus the pricing URL, or mark the table with an as-of date, so stale values don't silently skew mandatory cost estimates.

DimensionReasoningScore

Conciseness

The body is dense, operational, and free of explanations of concepts Claude already knows, but the card-binding/spending-limit warning appears twice nearly verbatim ("NEVER enter payment information… through Codex, Claude Code, or any CLI tool" and the closing "Cost protection" blockquote), cost-estimation guidance is duplicated across two sections, and exact $/sec prices are time-sensitive with no staleness marking — more than the minor trimming of anchor 4, but not the padded over-explanation of anchor 2.

3 / 5

Actionability

Patterns A–C are complete, executable Modal launchers with exact CLI commands and a concrete cost-estimation template, but Pattern D references undefined image/volume variables and Pattern E elides the core training code ("# ... transformers + peft + trl training code ..."), leaving minor gaps below the fully copy-paste-ready anchor 5 and well above the pseudocode-level anchor 3.

4 / 5

Workflow Clarity

A clearly sequenced 6-step workflow (Analyze → Estimate → Generate → Run → Verify → Collect/Cleanup) with real checkpoints — the mandatory "Before EVERY run, estimate cost and show to user for confirmation", an auth smoke test (`modal run -q 'print("ok")'`), and monitoring commands — but there is no error-recovery feedback loop (e.g., what to do on OOM, timeout, or failing logs), so it falls short of the anchor-5 validation-with-feedback-loops example.

4 / 5

Progressive Disclosure

Scored against the actual bundle structure: a single monolithic ~336-line SKILL.md with no references/, scripts/, or assets/ files, where the pricing table, benchmark comparison, six launcher patterns, and full CLI reference are all inlined; the sole reference (`../shared-references/compute-env-contract.md`) points outside the bundle. Section headers are well organized (above anchor 2's "minimal structure"), but bulk reference material that should be split out is inline, matching anchor 3.

3 / 5

Total

14

/

20

Passed

Description

96%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: concrete capabilities, an explicit 'Use when…' clause with abundant natural trigger phrases, and a clearly named platform niche. The only weakness is the generic "need a GPU" trigger, which risks firing when a non-Modal GPU platform is intended.

DimensionReasoningScore

Specificity

Lists four concrete workload actions ("training, fine-tuning, inference, batch processing") plus concrete platform mechanics ("no SSH, no Docker, auto scale-to-zero"), matching the anchor-5 pattern of multiple specific actions with comprehensive domain coverage; anchor 4 would require coverage gaps that aren't present.

5 / 5

Completeness

Explicitly answers both what ("Run GPU workloads on Modal — training, fine-tuning, inference, batch processing") and when ("Use when user says…") with concrete trigger phrases, exactly the anchor-5 shape; anchor 4 requires a less explicit 'when', which doesn't apply.

5 / 5

Trigger Term Quality

Eight natural trigger phrasings users would actually say — "modal run", "modal training", "deploy to modal", "need a GPU", "run on modal", "serverless GPU" — covering synonyms and quoted command forms, which matches the comprehensive anchor-5 example.

5 / 5

Distinctiveness Conflict Risk

Modal-specific triggers form a clear niche, but "need a GPU" and "serverless GPU" are generic and would fire for any GPU request, overlapping with closely related GPU-cloud skills the body itself names (vast.ai, run-experiment) — minor overlap risk per anchor 4 rather than the minimal-conflict anchor 5.

4 / 5

Total

19

/

20

Passed

Validation

87%

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

CriteriaDescriptionResult

allowed_tools_field

'allowed-tools' contains unusual tool name(s)

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

14

/

16

Passed

Repository

wanshuiyin/Auto-claude-code-research-in-sleep

Path

skills/skills-codex/serverless-modal/SKILL.md

Reviewed

Table of Contents

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.