The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a workflow for handling user-controlled chat messages via an LLM, creating a surface for indirect prompt injection.\n- Ingestion points: src/lib/ai/schemas.ts defines AiChatRequestSchema, which accepts a message string from users.\n- Boundary markers: The skill does not prescribe the use of delimiters (such as XML tags) or specific system instructions to ignore embedded commands in user input.\n- Capability inventory: The skill enables LLM invocation via the Vercel AI SDK within Next.js route handlers.\n- Sanitization: Input length is enforced via Zod, but no content-level sanitization or escaping is required before the input is passed to the AI provider.\n- [EXTERNAL_DOWNLOADS]: Fetches dependencies including the Vercel AI SDK (ai), zod, and official provider packages (@ai-sdk/openai, @ai-sdk/anthropic, etc.) from well-known registries via the Bun package manager.\n- [COMMAND_EXECUTION]: Executes local shell commands such as bun run lint, bun run build, test, and rg to verify code quality and project structure during integration.

addon-vercel-ai-labs