The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npm install and curl within its initialization scripts (scripts/init.sh, scripts/.prepare-template.sh) to download standard development dependencies from the NPM registry and to provision application credentials from a local platform portal. These are standard operations for development scaffolding tools and target well-known or internal platform services.
[COMMAND_EXECUTION]: The initialization process involves executing shell scripts and Node.js utilities to patch project configuration files (e.g., vite.config.ts, tsconfig.json, package.json). These operations are transparently documented in the SKILL.md and are essential for the skill's purpose of grafting a backend onto an existing frontend project.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by processing data from external sources (databases and API requests). However, the skill explicitly enforces Zod validation for all tRPC procedures as a mandatory best practice, which serves as an effective sanitization layer against malicious data payloads.
Ingestion points: Database tables (db/schema.ts) and tRPC procedure inputs (api/router.ts).
Boundary markers: tRPC input schemas using Zod.
Capability inventory: File system writes (scaffolding), database operations (Drizzle), and network requests (OAuth/API).
Sanitization: Mandatory use of Zod validation for all input objects.
[SAFE]: An automated security alert regarding the domain tsx.is was investigated. The domain belongs to the official tsx project (a widely used TypeScript execution engine). The reference in the skill is located within the package-lock.json metadata as a deprecation notice for an esbuild-related utility. This is a legitimate development resource and does not pose a security threat.

backend-building