Diagnoses wrong gradients in differentiable NVIDIA Warp programs by measuring first — comparing autodiff against finite differences on a shrunk reproduction before proposing any fix.
Cloudflare Sandbox SDKSkill
Summary
Build secure code-execution features on Cloudflare — sandboxes for AI-generated code, interpreters, and CI systems.
Features
- Secure execution of untrusted or model-generated code
- Sandbox lifecycle, filesystem, and process control
- Patterns for code interpreters and CI systems
Install This Skill
Add this skill to your favorite AI agent in a few steps.
Skill Content
Description
Any application that runs code it did not write needs an isolation boundary. Cloudflare's Sandbox SDK provides one, and this official skill covers using it correctly.
It walks through the sandbox lifecycle — creation, reuse, and teardown — along with filesystem access, process execution, streaming output back to a caller, and the resource limits that keep a runaway process from becoming a bill.
The patterns section covers the common builds: AI code interpreters, agent tool execution, CI/CD runners, and interactive development environments. Getting the boundary right matters more than usual here, because the code being run is by definition untrusted. Published by Cloudflare.
Related Skills
Google's official skill for the gws CLI — drive Gmail, Drive, Calendar, Sheets, Docs, Chat and Admin APIs from an agent, with Model Armor screening.
Netlify's official skill for zero-config managed Postgres — querying from Functions, Drizzle setup, migrations and per-preview database branches.
Official WordPress skill for Gutenberg block work: block.json, attributes and serialization, dynamic rendering, and the deprecation path that keeps existing content valid.
