The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands to manage deployments, including verifying the installation of the Vercel CLI with command -v vercel and executing the vercel deploy command.
[DATA_EXFILTRATION]: The scripts/deploy.sh script packages local project files into a tarball and uploads them to a Vercel-hosted deployment endpoint (https://codex-deploy-skills.vercel.sh/api/deploy). This is the intended primary behavior of the skill for facilitating cloud deployments.
The packaging process intentionally excludes sensitive directories and files, specifically .git, node_modules, .env, and .env.* patterns, reducing the risk of accidental credential or metadata exposure.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project files, which creates a potential surface for indirect prompt injection.
Ingestion points: The scripts/deploy.sh script reads and packages all files within the user-provided project directory (in scripts/deploy.sh).
Boundary markers: None identified for the file contents themselves; however, the skill follows a restricted execution flow that prioritizes preview deployments over production.
Capability inventory: Includes file system read access (tar), directory staging (mktemp, mkdir), and network transmission (curl).
Sanitization: The script implements explicit exclusion filters for environment files and version control history during the staging phase to sanitize the outbound payload.

vercel-deploy