The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted external data that could contain malicious instructions.
Ingestion points: The agent is instructed to read build and runtime logs via VERCEL_GET_DEPLOYMENT_LOGS and VERCEL_GET_RUNTIME_LOGS in SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided in SKILL.md to the agent to ignore potential instructions embedded within these logs.
Capability inventory: The skill documentation in SKILL.md describes significant write operations, including adding environment variables (VERCEL_ADD_ENVIRONMENT_VARIABLE), creating DNS records (VERCEL_CREATE_DNS_RECORD), and triggering deployments (VERCEL_CREATE_NEW_DEPLOYMENT).
Sanitization: There are no instructions in SKILL.md for the agent to sanitize or validate the content of the logs before using that information to perform subsequent actions.
[EXTERNAL_DOWNLOADS]: The skill requires connection to an external MCP server to fetch tool schemas and perform platform operations.
Evidence: Instructions in SKILL.md direct users to add https://rube.app/mcp as an MCP server. This endpoint provides the underlying tool definition for the Vercel automation tasks.

vercel-automation