sandbox-sdk

Build sandboxed applications for secure code execution. Load when building AI code execution, code interpreters, CI/CD systems, interactive dev environments, or executing untrusted code. Covers Sandbox SDK lifecycle, commands, files, code interpreter, and preview URLs. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

2.63x

Quality

93%

Does it follow best practices?

Impact

100%

2.63x

Average score across 3 eval scenarios

Securityby

Passed

No findings from the security scan

The canonical home for this skill is sandbox-sdk in cloudflare/skills

Quality

Content

87%Weight 40%Scale 1-3

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is lean, highly actionable, and well-structured with proper progressive disclosure to real reference files. Its weak spot is workflow clarity: it opens with a verification checkpoint but the configure/deploy path lacks later validation steps or a feedback loop.

Suggestions

Add a short numbered deploy workflow after "Required Configuration" — e.g. configure wrangler.jsonc → add the Sandbox export → run `npx wrangler deploy` → verify the container binding responds — so the multi-step path is explicitly sequenced.

Add a validation/checkpoint step after deploy (such as confirming `wrangler deploy` succeeds and the Sandbox binding is live) with a "if it fails, check the Sandbox export and migrations" feedback loop.

Move or annotate the pinned base image version (`docker.io/cloudflare/sandbox:0.7.0`) so it is clearly flagged as version-sensitive, since hard-coded version numbers risk staleness over time.

Dimension	Reasoning	Score
Conciseness	The body is lean — terse tables, runnable code blocks, and short notes like "Keep images lean - affects cold start time" — with no padding explaining concepts Claude already knows. It assumes competence and every section earns its place, matching the level-3 anchor.	3 / 3
Actionability	It provides copy-paste-ready, executable guidance throughout: `npm install`, a full wrangler.jsonc, the required Worker export, a Quick Reference table of concrete method calls, and complete TypeScript/Dockerfile examples. This matches the level-3 anchor of fully executable code with specific examples.	3 / 3
Workflow Clarity	"FIRST: Verify Installation" gives one explicit checkpoint (`docker info # Must succeed`), but the broader configure→deploy workflow (wrangler.jsonc, Sandbox export, deploy) is reference-style with no subsequent validation or feedback loop. The single checkpoint is not enough to reach level 3, and it is above level 1 because a sequence and one checkpoint are present.	2 / 3
Progressive Disclosure	The body is a well-organized overview (quick reference, core patterns, lifecycle, anti-patterns) that defers full detail to one-level-deep, clearly signaled references — `references/api-quick-ref.md` ("Full API with options and return types") and `references/examples.md` ("Example index with use cases"), both of which exist. Content that should be separate is separate, matching the level-3 anchor.	3 / 3
	Total	11 / 12 Passed

Description

100%Weight 40%Scale 1-3

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 distinctive: it names concrete capabilities, gives an explicit "Load when..." trigger with natural user terms, and is clearly scoped to the Cloudflare Sandbox SDK. It avoids vague fluff and uses third-person/imperative voice with no first/second-person slip.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities — "Build sandboxed applications for secure code execution" plus "Sandbox SDK lifecycle, commands, files, code interpreter, and preview URLs" — rather than vague language. It is comprehensive, not just naming a domain, so it clears the level-2 anchor.	3 / 3
Completeness	It states what ("Build sandboxed applications... Covers Sandbox SDK lifecycle, commands, files, code interpreter, and preview URLs") and an explicit when ("Load when building AI code execution, code interpreters, CI/CD systems... or executing untrusted code"). Both halves are present with an explicit trigger clause, satisfying the level-3 anchor.	3 / 3
Trigger Term Quality	"building AI code execution, code interpreters, CI/CD systems, interactive dev environments, or executing untrusted code" covers the natural phrases a user would say when they need this skill. It is not jargon-only and covers common variations, matching the level-3 anchor.	3 / 3
Distinctiveness Conflict Risk	The Cloudflare Sandbox SDK is a clear niche and "Biases towards retrieval from Cloudflare docs over pre-trained knowledge" pins it to Cloudflare, with distinct triggers unlikely to fire for unrelated skills. It does not read as generic, matching the level-3 anchor.	3 / 3
	Total	12 / 12 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: openai/plugins
Path: plugins/cloudflare/skills/sandbox-sdk/SKILL.md
Commit: 11c74d6

Reviewed: 17 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.