meme-trench-scanner

Fail

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The local dashboard UI (assets/dashboard.html) exhibits an indirect prompt injection surface. Ingestion points: Token symbols, names, and transaction messages retrieved from the blockchain via onchainos. Boundary markers: Absent; external data is rendered directly into the UI. Capability inventory: Transaction signing and execution via onchainos CLI. Sanitization: None; the dashboard uses innerHTML to render token metadata, allowing a maliciously named token to potentially execute arbitrary JavaScript (XSS) in the user's browser.
  • [COMMAND_EXECUTION]: The script scan_live.py utilizes subprocess.run with shell=True to clear the dashboard port during startup. While this is limited to a local environment management task, using the shell to execute concatenated strings is a security anti-pattern.
  • [EXTERNAL_DOWNLOADS]: The bot relies on the onchainos CLI, which is an external dependency downloaded from onchainos.com. The skill provides the installation command (curl | bash) to the user via documentation and console output. While this is a common installation method for developers, it involves executing unverified remote scripts.
Recommendations
  • HIGH: Downloads and executes remote code from: https://onchainos.com/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 4, 2026, 09:25 AM