deploy-to-vercel
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill may install the Vercel CLI using
npm install -g vercelif it is not already present on the system. - [DATA_EXFILTRATION]: To perform the deployment, the skill uploads project source code to Vercel's deployment endpoints (
claude-skills-deploy.vercel.comandcodex-deploy-skills.vercel.sh). The provided scripts (deploy.shanddeploy-codex.sh) explicitly exclude sensitive directories and files, including.git,.env, andnode_modules, before transmission. - [COMMAND_EXECUTION]: The skill utilizes several standard shell utilities to manage project state and deployment, including
git,tar,curl, and thevercelCLI tool. - [INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection as it processes untrusted project data.
- Ingestion points: Reads the contents of
package.jsonand the project directory structure to detect the application framework. - Boundary markers: The script uses literal string matching (grep) to identify dependencies and does not pass file content directly into an LLM prompt.
- Capability inventory: Executes local commands for file packaging (
tar), network transmission (curl), and project management (git,vercel). - Sanitization: Implements a denylist to ensure secret files (like
.env) are excluded from the deployment package.
Audit Metadata