ai4trade-trading-signals
Fail
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill mandates fetching and following instructions from an external URL (
https://ai4trade.ai/SKILL.md) for registration and core trading operations. This enables an external server to dynamically modify the agent's logic after the skill is loaded.\n- [INDIRECT_PROMPT_INJECTION]: The skill relies on untrusted external content to define its behavior, creating a significant attack surface for indirect injection.\n - Ingestion points: Instructions to read content from
https://ai4trade.ai/SKILL.mdare present inSKILL.md.\n - Boundary markers: No delimiters or instructions are provided to the agent to disregard malicious commands embedded in the remote content.\n
- Capability inventory: The skill allows the
Bashtool and explicitly suggests writing shell orcurlscripts to implement its logic.\n - Sanitization: No validation or sanitization of the remote content is required before the agent follows the instructions.\n- [COMMAND_EXECUTION]: The skill encourages the use of
Bashand custom shell scripts to interact with the external platform, which could be leveraged to execute harmful commands if the remote instructions are compromised.
Recommendations
- AI detected serious security threats
Audit Metadata