money-discover
Pass
Audited by Gen Agent Trust Hub on May 16, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill instructions are focused exclusively on business analysis workflows and do not contain patterns for bypassing safety filters, extracting system prompts, or overriding agent constraints.
- [DATA_EXFILTRATION]: No evidence of hardcoded credentials or access to sensitive local file paths (such as .ssh or .env) was found. The skill collects user-provided business context but does not exfiltrate this data to unauthorized external endpoints.
- [REMOTE_CODE_EXECUTION]: The skill does not perform any external downloads, package installations, or dynamic code execution. It relies on standard search and reasoning capabilities.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources during its 'Opportunity Scanning' phase (e.g., searching Reddit, Product Hunt, and Hacker News). While this constitutes an attack surface, the risk is mitigated as the skill uses this information for brainstorming and does not execute commands based on the external content.
- Ingestion points: Phase 2 (Trend Analysis and Problem Mining) in SKILL.md fetches data from external web sources.
- Boundary markers: None explicitly implemented to isolate search results.
- Capability inventory: Browsing public web content and performing competitive research.
- Sanitization: Not explicitly defined in the prompt instructions.
Audit Metadata