benchmark-sandbox
Run vercel-plugin eval scenarios in Vercel Sandboxes instead of local WezTerm panels. Provisions ephemeral microVMs with Claude Code + plugin pre-installed, runs benchmark prompts, extracts hook artifacts, and produces coverage reports.
78
68%
Does it follow best practices?
Impact
92%
2.09xAverage score across 3 eval scenarios
Medium
Suggest reviewing before use
Fix and improve this skill with Tessl
tessl review fix ./.claude/skills/benchmark-sandbox/SKILL.mdSecurity
1 medium severity finding. This skill can be installed but you should review these findings before use.
W013: Attempt to modify system services in skill instructions.
What this means
The skill prompts the agent to compromise the security or integrity of the user’s machine by modifying system-level services or configurations, such as obtaining elevated privileges, altering startup scripts, or changing system-wide settings.
Why it was flagged
The skill instructs agents to run Claude with the --dangerously-skip-permissions flag (explicitly bypassing permission prompts) and to write CLI auth tokens into user auth files, which encourages bypassing security controls on the host sandbox even though it does not request sudo, edit system files, or create users.
Low
Low-risk findings.
2 low severity findings. Worth noting, but not necessarily harmful.
W011: Third-party content exposure detected (indirect prompt injection risk).
What this means
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
Why it was flagged
In `run-eval.ts`, a scenario’s outsider-authored free text prompt (from `--scenarios-file`) is written into `/tmp/prompt.txt` and then ingested by Claude Code via `claude ... "$(cat /tmp/prompt.txt)"` in the BUILD and VERIFY/DEPLOY phase prompts.
W012: Unverifiable external dependency detected (runtime URL that controls agent).
What this means
The skill fetches instructions or code from an external URL at runtime, and the fetched content directly controls the agent’s prompts or executes code. This dynamic dependency allows the external source to modify the agent’s behavior without any changes to the skill itself.
Why it was flagged
The spike/interactive-proof.ts script executes `npx -y add-plugin https://github.com/vercel/vercel-plugin` at runtime (line spike/interactive-proof.ts:139), which fetches and installs remote code from that GitHub URL into the sandbox and thus executes/controls plugin behavior and hooks.
- Repository
- vercel/vercel-plugin
- Commit
358400a
- Audited
- Security analysis
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.