Smart Contract Audit

Warn

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill's install.command and setup instructions perform a remote fetch using curl -s https://ethskills.com/audit/SKILL.md. This downloads instruction sets from an external domain not associated with the skill's local repository.
  • [REMOTE_CODE_EXECUTION]: The setup instructions in catalog.json recommend installing external code through multiple channels:
  • claude plugin install https://github.com/austintgriffith/ethskills installs a plugin from a third-party personal GitHub repository.
  • clawhub install ethskills suggests installing a package via an external tool manager.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data provided by the user in the form of smart contract code.
  • Ingestion points: The skill explicitly asks the user to "Audit this contract: " (identified in catalog.json).
  • Boundary markers: There are no specific instructions provided to the agent to treat the pasted Solidity code as untrusted or to ignore embedded instructions within comments in that code.
  • Capability inventory: The skill description and demo indicate the ability to "File GitHub issues for anything high or critical," which involves network write operations.
  • Sanitization: No sanitization or validation of the input Solidity code is described before the agent processes it.
  • Risk: Maliciously crafted smart contracts (e.g., using comments containing agent instructions) could attempt to influence the agent to file fake issues, exfiltrate data through issue descriptions, or perform other unintended GitHub operations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 1, 2026, 10:16 PM