moon-dev-trading-agents
Fail
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: HIGHDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documentation for rbi_agent.py states that the agent "Generates backtesting.py compatible code" and then "Executes backtest". Performing autonomous code generation and execution at runtime from instructions extracted from untrusted external sources is a high-risk capability that can lead to local code execution.\n- [INDIRECT_PROMPT_INJECTION]: The rbi_agent.py component is designed to process data from untrusted external sources, including YouTube URLs and PDFs. This ingestion surface is vulnerable to indirect prompt injection.\n
- Ingestion points: The rbi_agent.py script processes YouTube URLs, PDFs, and trading idea text as described in SKILL.md.\n
- Boundary markers: There are no mentioned delimiters or instructions to ignore embedded commands within the ingested data.\n
- Capability inventory: The skill environment has capabilities for file writing, network access (via exchange APIs), and the execution of generated Python scripts.\n
- Sanitization: No sanitization, validation, or human-review steps are described for the external content before it is converted into executable code.
Recommendations
- AI detected serious security threats
Audit Metadata