evm-deep-dive
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION] (SAFE): No attempts to override agent constraints or bypass safety filters. The identity and role-play instructions are strictly relevant to the technical domain of blockchain engineering.
- [DATA_EXFILTRATION] (SAFE): No network operations (curl, wget, fetch) or access to sensitive local file paths (~/.ssh, ~/.aws) were identified. All code snippets are self-contained Solidity/Yul examples.
- [REMOTE_CODE_EXECUTION] (SAFE): No external script downloads or execution patterns. The skill does not perform any package installations (npm, pip) or use dynamic evaluation (eval, exec) within the host environment.
- [INDIRECT_PROMPT_INJECTION] (LOW): The skill is designed to analyze user-provided Solidity code, which is a potential surface for indirect injection via comments. However, the skill lacks the capabilities (network, file-write, or shell access) to be exploited by such an injection. The risk is limited to the agent providing biased optimization advice.
- [OBFUSCATION] (SAFE): No Base64, zero-width characters, or homoglyph-based obfuscation detected. The assembly bytecode provided in the EIP-1167 pattern is standard and well-documented for its stated purpose.
Audit Metadata