dcg

Fail

Audited by Gen Agent Trust Hub on Jun 3, 2026

Risk Level: HIGHPROMPT_INJECTIONREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The file AGENTS.md includes an 'Override Prerogative' (Rule 0) that explicitly commands the AI to disregard its existing constraints and filters in favor of the author's instructions.
  • [REMOTE_CODE_EXECUTION]: The documentation and skill files promote an installation method that involves fetching and executing a remote shell script from GitHub (install.sh) directly into the shell.
  • [COMMAND_EXECUTION]: The dcg setup command is designed to modify system shell initialization files (~/.bashrc, ~/.zshrc) to implement persistent startup checks, a pattern common in persistence mechanisms.
  • [COMMAND_EXECUTION]: The skill utilizes a persistent local SQLite database (fsqlite) to log all command execution history, which could expose a log of sensitive user activities within the agent's environment.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/Dicklesworthstone/destructive_command_guard/main/install.sh?$(date, https://sh.rustup.rs, https://raw.githubusercontent.com/Dicklesworthstone/destructive_command_guard/main/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 3, 2026, 02:01 AM
Security Audit — agent-trust-hub — dcg