The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents patterns for processing untrusted user input through SvelteKit form actions and API endpoints. It mitigates this risk by explicitly mandating server-side validation using schemas (Zod or Valibot) and enforcing authentication checks in server-side logic.
Inestion points: User-submitted form data via request.formData() and the extractFormData utility in references/forms-validation.md.
Boundary markers: Instructions consistently advocate for validating all external data and verifying locals.session before processing.
Capability inventory: Database interactions (Drizzle/D1) and network operations.
Sanitization: Emphasizes the use of structural validation and returning field-keyed errors via fail().
[COMMAND_EXECUTION]: The references/cloudflare.md file includes an example developer script (scripts/create-user.ts) that uses execSync to run wrangler CLI commands. This is identified as standard developer tooling for managing Cloudflare D1 databases and does not represent an automated or malicious execution pattern.

sveltekit