scroll-world-storytelling
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted source material (articles, case studies, or narratives) to generate landing page structures, copy, and specific prompts for asset generation. This workflow creates a vulnerability surface for indirect prompt injection.
- Ingestion points: The 'Read the whole source' section in
SKILL.mdexplicitly directs the agent to ingest external articles or narratives provided by users. - Boundary markers: The instructions lack defined delimiters or specific 'ignore' directives to separate the source material's data from potential instructions embedded within that data.
- Capability inventory: The skill possesses significant capabilities, including the generation of functional HTML and JavaScript (specifically Three.js) and the creation of prompts for external media generation tools (Grok Imagine).
- Sanitization: No input validation, escaping, or filtering mechanisms are described to mitigate risks from maliciously crafted source text.
Audit Metadata