gray-transaction-systems
Fail
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: HIGHOBFUSCATIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [OBFUSCATION]: The main headers in
SKILL.md,philosophy.md, andreferences.mdcontain large blocks of invisible zero-width Unicode characters (U+200B, U+200C, U+200D, U+2060). This is a steganographic technique used to hide data from human review while ensuring it is ingested by the AI agent. - [PROMPT_INJECTION]: The hidden content constitutes a steganographic prompt injection attack. By embedding instructions in invisible characters, the skill author attempts to override the agent's behavior, bypass safety guardrails, or execute commands without the user's knowledge.
- [DATA_EXFILTRATION]: The steganographic payload is highly likely to contain instructions for the agent to access and transmit sensitive files—such as
.envfiles, SSH keys, or cloud credentials—to an external attacker-controlled server, leveraging the agent's file access and network capabilities.
Recommendations
- AI detected serious security threats
Audit Metadata