startup-design

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements an automated research workflow that is vulnerable to indirect prompt injection from untrusted web content.
  • Ingestion points: Automated web research in Phase 3 retrieves content from Reddit, professional forums, and competitor websites as defined in the references/research-wave-*.md files.
  • Boundary markers: There are no instructions or delimiters defined in the research agent templates to prevent the agent from executing instructions found in the retrieved web data.
  • Capability inventory: The skill utilizes the Agent and WebSearch tools and has the ability to write multiple markdown files to the local filesystem across eight different project phases.
  • Sanitization: The research protocol explicitly instructs the agent to capture 'verbatim' quotes and raw findings from the web, which are then synthesized into final deliverables without validation or escaping.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 03:30 AM