The Agent Skills Directory

Indirect Prompt Injection Surface: The skill includes instructions to scan project files to identify API endpoints and suggest rate-limiting rules. This process involves the ingestion of potentially untrusted data from the local repository, which could influence the agent's logic if the files contain malicious instructions. * Ingestion points: Local project files are scanned to identify and define API structures (SKILL.md). * Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore instructions found within the scanned files. * Capability inventory: The skill leverages the vercel CLI to perform mutations on the project's firewall configuration. * Sanitization: The instructions do not specify sanitization or validation routines for the data extracted from repository files.
Restricted Command Access: The skill proactively identifies and blocks high-risk or interactive commands (such as Attack Mode and pausing system mitigations) from being executed by the agent, which is a positive security design that preserves user control over critical security states.

vercel-firewall