viral-generator-builder

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection (LOW): The skill encourages the creation of tools that ingest untrusted user data (names, birthdays, quiz answers, and photo uploads) to be used as input for 'AI-powered' or algorithmic generation. This creates a potential attack surface if the agent handles this data without boundaries.
  • Ingestion points: SKILL.md (Input Design section, Question Design section).
  • Boundary markers: Absent; the provided patterns do not include delimiters or instructions to the agent to disregard embedded instructions within user-provided inputs.
  • Capability inventory: The skill itself contains static JavaScript logic for hashing and scoring; it does not invoke dangerous tool capabilities, file system operations, or network requests.
  • Sanitization: Absent; the architectural templates lack validation or escaping logic for the untrusted data being processed.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 04:47 AM