blockchain-privacy
Fail
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: CRITICALEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The reference file
references/sharp_edges.mdcontains a link tohttps://tornado.ws/compliance/, which has been flagged by security scanners as a phishing and malicious site. This domain is a suspicious mirror of the officialtornado.cashprotocol and could lead users to interact with a compromised version of the service. - [INDIRECT_PROMPT_INJECTION]: The skill acts as an expert researcher and processes domain-specific information about blockchain privacy protocols. While no active injection was found, the skill's structure creates a surface where malicious instructions could be embedded in the technical data it processes.
- Ingestion points: Technical queries and documentation related to blockchain privacy and ZK protocols.
- Boundary markers: Absent. There are no explicit instructions for the agent to ignore or delimit embedded instructions in the processed reference material.
- Capability inventory: The skill provides complex smart contract code (Solidity, Circom) and suggests usage of external RPC providers and relayers, which are high-leverage actions.
- Sanitization: Absent. No evidence of input validation or escaping for the technical content being analyzed.
Recommendations
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata