The Agent Skills Directory

[Indirect Prompt Injection]: The skill processes feature flag configuration data (such as keys, descriptions, and variants) that can be fetched from remote providers or shared projects. This represents a potential surface where external content could influence the agent's context.
Ingestion points: Data enters through CLI commands like vercel flags inspect and via the Flags Explorer discovery endpoint.
Capability inventory: The skill allows the agent to execute CLI commands for project management and environment variable configuration.
Boundary markers: There are no specific instructions for the agent to treat flag metadata as untrusted or to use delimiters when processing descriptions.
Sanitization: The skill mentions safeJsonStringify for protecting browser-side data, but server-side handling of flag data depends on the agent's own safeguards.
[Command Execution]: The skill makes extensive use of the Vercel CLI to manage project state and pull sensitive credentials.
Evidence: Instructions include commands such as vercel link, vercel env pull, and vercel flags create to perform administrative tasks.
Context: These operations are performed using official tools and follow documented workflows for managing feature flags in the Vercel ecosystem.

flags-sdk