The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION] (LOW): The skill implements logic that processes untrusted external data (HTTP request headers and bodies) via authentication handlers.
Ingestion points: The auth.handler(c.req.raw) method in references/cloudflare-worker-kysely.ts and references/nextjs/postgres-example.ts ingests raw request data.
Boundary markers: Absent. There are no specific instructions provided to the agent to ignore instructions embedded within the user data processed by these handlers.
Capability inventory: The skill provides logic for session verification and database queries but does not include dangerous capabilities like arbitrary command execution or file system writes.
Sanitization: Relies on the internal validation and sanitization logic of the third-party better-auth library.
[DATA_EXPOSURE & EXFILTRATION] (SAFE): Code examples correctly utilize environment variable bindings (e.g., env.BETTER_AUTH_SECRET) rather than hardcoding credentials. Database connection strings in documentation use generic placeholders.
[REMOTE_CODE_EXECUTION] (SAFE): No patterns for remote script execution or piped downloads were detected. The skill references standard, well-known libraries in the TypeScript ecosystem.

better-auth