About modal-sandboxes-knowledge
A reference skill for designing, securing, and operating Modal Sandboxes. It focuses on untrusted code execution, resource and network limits, lifecycle cleanup, and audit-friendly output handling.
Key features
- 01Security checklist for untrusted inputs
- 02CPU, memory, disk, and timeout limits
- 03Deny-by-default network egress guidance
- 04Sandbox lifecycle and cleanup patterns
- 05Separate stdout, stderr, and return codes
Use cases
- 01Running submitted code in isolated sandboxes
- 02Restricting sandbox network access
- 03Reusing session sandboxes for interactive work
- 04Auditing command output and exit status
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/modal-sandboxes-knowledgein Claude Code to confirm it loaded.
Bundled filesreferences/sandboxes-code-execution.md
Details
From the project READMEModal Sandboxes Knowledge
Use this skill for Modal Sandbox design, security, and operations. For ordinary functions, GPU jobs, or deployment topics, use modal-compute:modal-compute-knowledge.
Essential security checklist
- Treat all submitted code and files as untrusted.
- Use explicit timeout, CPU, memory, and disk constraints.
- Prefer deny-by-default network egress; allow only required destinations.
- Clean up every sandbox with
terminate()or a scoped lifecycle. - Avoid passing long-lived credentials into sandbox environments unless strictly required.
- Capture stdout/stderr and return codes separately for auditing and debugging.
Common patterns
- One-shot execution: create a sandbox, run a command, collect output, terminate.
- Named sandbox: reuse a per-session sandbox for interactive workflows.
- PTY session: support shell-like interaction while preserving resource limits.
- Snapshot workflow: prebuild filesystem state, then launch from a known baseline.
Detailed reference
See references/sandboxes-code-execution.md for code examples and operational notes.
Comments
Markdown supported