The Agent Skills Directory

[SAFE]: The instructions promote secure environment variable management by guiding users to generate unique secrets using openssl rather than hardcoding values.
[SAFE]: External dependencies are restricted to well-known packages in the React and Convex ecosystems, specifically better-auth and @convex-dev/better-auth.
[SAFE]: Implementation patterns for sign-up and sign-in flows utilize standard client-side libraries that handle sensitive data through established protocols.
[INDIRECT_PROMPT_INJECTION]: The skill establishes patterns for handling untrusted data at authentication ingestion points.
Ingestion points: User input is ingested via authClient.signUp.email and authClient.signIn.email in references/setup-guide.md.
Boundary markers: Standard library abstractions are used, though specific prompt-level delimiters for the agent are not defined in the code examples.
Capability inventory: The skill uses fetchAuthMutation and fetchAuthQuery in src/lib/auth-server.ts to perform authenticated server operations.
Sanitization: Validation and sanitization are delegated to the better-auth framework and Convex backend logic.

betterauth-tanstack-convex