The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill instructs the agent to execute a wide range of powerful system commands, including vercel deploy, vercel env pull, and npm run build. These commands can modify project states, retrieve environment variables, and execute arbitrary build scripts.
[EXTERNAL_DOWNLOADS] (LOW): The skill recommends installing the vercel CLI via npm install -g vercel. Per [TRUST-SCOPE-RULE], this is downgraded to LOW as Vercel is a recognized and trusted deployment platform.
[Indirect Prompt Injection] (MEDIUM): The skill is intended for 'troubleshooting deployment issues', which requires the agent to ingest and reason about external data such as build logs and error messages.
Ingestion points: Vercel build logs, CLI error outputs, and dashboard logs referenced in the 'Troubleshooting' and 'Debug steps' sections.
Boundary markers: None identified. There are no instructions to the agent to treat external logs as untrusted data.
Capability inventory: The skill provides full access to the Vercel CLI (vercel), including deployment, environment variable management, and project linking.
Sanitization: None. The skill assumes logs are informative and does not suggest sanitizing log content before the agent interprets it for next steps.
[DATA_EXPOSURE] (LOW): The skill explains how to access sensitive files like .vercel/project.json and .env.local. While these are legitimate parts of the Vercel workflow, an agent with this skill has the capability to read and potentially expose project IDs and local secrets.

adynato-vercel