founder-brand-onboarding
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill collects a significant amount of untrusted user data (voice patterns, stories, audience definitions) and writes it directly to
SOUL.mdandbrand-config.json. This content is designed to be read and used as context for future content generation, which creates a potential vulnerability where malicious instructions embedded in the user data could influence the agent's future behavior. - Ingestion points: Interactive user responses throughout the 10-step onboarding process, specifically in steps collecting phrases, voice settings, and story reservoirs.
- Boundary markers: The instructions lack specific guidance on using delimiters or XML-style tagging to separate user-provided content from agent instructions in the generated configuration files.
- Capability inventory: The skill uses
Writeto create local configuration files andRead/Grepto access them, allowing user-sourced data to enter the agent's reasoning loop. - Sanitization: No validation or escaping of user input is mentioned before the data is written to the filesystem.
Audit Metadata