new-app

Pass

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Use of a hardcoded absolute file path to a specific user's home directory ('/Users/andrzej/.rbenv/versions/3.4.6/bin/rails') for command execution.\n- [COMMAND_EXECUTION]: Instruction to execute 'rm -rf' to remove a nested .git directory within the newly created application path.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface.\n
  • Ingestion points: User-provided values for '{app_name}', '{domain_name}', and '{port}' are interpolated into commands.\n
  • Boundary markers: Absent; no separators used to isolate user-provided data from shell instructions.\n
  • Capability inventory: High-privilege shell access via 'rails', 'bundle', 'rm', and 'make' commands.\n
  • Sanitization: No validation or escaping of input placeholders is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 20, 2026, 02:16 PM
Security Audit — agent-trust-hub — new-app