A Vercel Sandbox Claude sends work to on purpose: long test runs in the background, someone else's repository, an installer of unknown origin, a bulk change to preview as a patch, a clean build. TypeSafe's Jev reads each Bash command and spots those jobs; by default you are asked whether it runs in the sandbox or locally (confirmSandbox), or it always goes to the sandbox with the question off. Claude also gets sandbox_run, sandbox_jobs, sandbox_result and sandbox_apply. Each job starts from the project as it is now (tracked files, credentials left out) in its own worktree, every result is…
- Stars
- 33,101
- Version
- v1
- Copies
- 0
Overview
Waxell rates it high risk: 5 findings, highest severity critical.
- Highest severity
- critical
- Findings
- 5
- What ran
- Static analysis
A clean result means nothing alarming in what is published, not a guarantee. How the checks work
We don't republish the plugin's instructions. Read them on GitHub at the reviewed commit.
Security
Security scan
Checked 2 days ago
Result
High risk Waxell rates it high risk: 5 findings, highest severity critical.
Checked with static analysis: servers it launches, packages it installs at runtime, scripts piped to a shell and bundled binaries. Nothing was run. How the checks work
Findings (5)
critical · 4
- critical
Tells Claude to download a script and pipe it to a shell
README.md:15 · skill_risk:S1
- critical
Tells Claude to download a script and pipe it to a shell
hooks/judge.ts:90 · skill_risk:S1
- critical
Tells Claude to download a script and pipe it to a shell
tests/jev-vercel-sandbox.test.tsx:344 · skill_risk:S1
- critical
Tells Claude to download a script and pipe it to a shell
tests/jev-vercel-sandbox.test.tsx:423 · skill_risk:S1
high · 1
- high
Tells Claude to install unpinned packages (a, docker-free, release, build)
hooks/judge.ts:102 · skill_risk:S2
Versions
Version history
Each distinct content we have seen, newest first, with the verdict it got at the time.
One version so far: v1 (0.3.0), seen 2026-11-03.
Files
Files
10 files, 161 KB in total. Names only; open them on GitHub.
- .claude-plugin/plugin.json
- README.md
- hooks/hooks.json
- hooks/jev-vercel-sandbox.tsx
- hooks/judge.ts
- hooks/vercel.ts
- hooks/workspace.ts
- tests/jev-vercel-sandbox.test.tsx
- tests/judge.test.ts
- tests/workspace.test.ts
See and control which skills your people use
Inside your organization, Waxell scans each skill and plugin before anyone installs it, approves what passes and publishes it to a private marketplace, and puts every MCP tool call behind one governed URL.
30 minutes. No commitment.
SOC 2 Type II attested · ISO/IEC 27001:2022 certified · HIPAA with a BAA · aligned to the EU AI Act and NIST AI RMF
More engineering skills and plugins
fine-tuning-axolotldavila7Expert guidance for fine-tuning LLMs with Axolotl - YAML configs, 100+ models, LoRA/QLoRA, DPO/KTO/ORPO/GRPO, multimodal support High riskEngineeringDocuments33k
fine-tuning-unslothdavila7Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization High riskEngineeringDocuments33k
infrastructure-lambda-labsdavila7Reserved and on-demand GPU cloud instances for ML training and inference. Use when you need dedicated GPU instances with simple SSH… High riskEngineeringAI & agents33k
using-neondavila7Guides and best practices for working with Neon Serverless Postgres. Covers getting started, local development with Neon, choosing a… High riskEngineeringData & analytics33k