The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted external data through route parameters and search parameters which could be used as a vector for injection if processed by an LLM downstream.
Ingestion points: External data enters the context via dynamic route parameters (params.userId in app/routes/users/$userId.tsx), search parameters in app/routes/users/index.tsx, and user-provided form data in the Users component.
Boundary markers: The skill does not explicitly use prompt-specific boundary markers, but relies on standard application logic boundaries.
Capability inventory: The skill performs network requests via fetch and database writes/reads using bun:sqlite in server functions and API routes.
Sanitization: The skill demonstrates good security practices by using zod for search parameter validation and parameterized SQL queries (db.run(..., [name])) to prevent SQL injection.
[EXTERNAL_DOWNLOADS]: The skill instructs users to use scaffolding tools that fetch and execute remote code during project initialization.
Evidence: The Quick Start section utilizes bunx create-tanstack-start@latest, which downloads the latest version of the scaffolding tool from the NPM registry.
Context: The skill includes a dedicated "Secure Installation" section that explicitly warns users about supply chain risks and recommends mitigations such as using trustedDependencies, minimumReleaseAge, and package auditing tools like Socket.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands to manage the project lifecycle.
Evidence: Commands include bun install, bun run dev, bun run build, and bun run start for local development and production builds.

bun-tanstack-start