securityterminal

Terminal Security

Terminal emulators interpret escape sequences that can be weaponized. LLM responses are untrusted.

ikigai Application

ANSI escape injection:

• LLM responses may contain malicious escape sequences
• Attackers can: change terminal title, redefine keys, write to arbitrary screen positions
• Some terminals vulnerable to escape sequences that execute commands

Sanitization strategy:

• Strip or escape control characters (0x00-0x1F, 0x7F) except safe ones
• Allowlist: \n, \r, \t for formatting
• Escape or strip: \x1b (ESC), \x9b (CSI)
• Consider: pass through only after validation

Safe display patterns:

// Filter control chars before display
for (size_t i = 0; i < len; i++) {
    if (data[i] < 0x20 && data[i] != '\n' && data[i] != '\t') {
        continue;  // Skip control char
    }
    output_char(data[i]);
}

Raw mode considerations:

• Restore terminal state on exit (even on crash)
• Handle SIGINT/SIGTERM to cleanup
• Don't leak raw mode to child processes

Review red flags: Unsanitized LLM output to terminal, missing escape filtering, no terminal cleanup on error paths.