backend-patterns

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from multiple untrusted sources, creating an attack surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the agent context through formData in Server Actions within app/actions/users.ts, raw request bodies in Supabase Edge Functions in supabase/functions/process-webhook/index.ts, and event data payloads in Inngest background functions in inngest/functions.ts.
  • Boundary markers: The instructions advise using Zod validation and session-based authentication checks as logical boundaries to isolate external content processing.
  • Capability inventory: The skill includes instructions for database mutations using Prisma and Supabase clients, as well as network operations via fetch and event dispatching tools.
  • Sanitization: The skill explicitly requires the use of Zod schemas to validate and sanitize all external inputs prior to any execution or storage logic.
  • [COMMAND_EXECUTION]: The skill instructs the agent to run discovery commands to audit the existing environment.
  • In the Check existing first section of SKILL.md, the agent is directed to use ls, cat, and rg to inspect project structure and configuration files like package.json and prisma/schema.prisma. These are benign discovery operations intended for architectural analysis.
  • [EXTERNAL_DOWNLOADS]: The skill references and integrates libraries from well-known service providers and public CDNs.
  • It imports the Supabase JavaScript SDK from the esm.sh content delivery network for Edge Function implementation.
  • It utilizes established Node.js packages such as @upstash/ratelimit, @upstash/redis, and inngest for essential backend functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:26 PM