The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill makes outbound network requests using curl to the endpoint https://codex-deploy-skills.vercel.sh/api/deploy to upload project files and poll the deployment status.
[DATA_EXFILTRATION]: The scripts/deploy.sh script packages the project directory into a tarball and uploads it to an external deployment service. The script minimizes risk by explicitly excluding sensitive directories and files, including node_modules, .git, .env, and .env.* files, preventing the accidental transmission of secrets.
[COMMAND_EXECUTION]: The skill utilizes several standard command-line tools such as tar for archiving, grep and sed for framework detection and JSON parsing, and curl for network communication. It also provides instructions for the agent to use the vercel CLI if available.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled files, specifically package.json, to automatically detect the project framework and configure deployment settings.
Ingestion points: The scripts/deploy.sh script reads the content of package.json from the project directory.
Boundary markers: The skill defines specific JSON output formats for the deployment service response to maintain structured communication.
Capability inventory: The skill has capabilities for file reading, archive creation, directory traversal, and network transmission.
Sanitization: The skill uses explicit exclusion flags in the tar command to filter out sensitive project metadata and credential files before they are processed or uploaded.

vercel-deploy