Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Security: aezizhu/g

Security

docs/SECURITY.md

Security Model

Author: aezizhu

Threat Model

  • Accidental harmful commands due to misinterpretation
  • Malicious prompt injections
  • Privilege escalation or environment abuse

Controls

  • Shell-free execution: commands are argv arrays, no pipelines or redirections
  • Allowlist and denylist regexes checked against entire command line
  • Minimal environment: only PATH preserved
  • Per-command timeouts; SIGTERM then SIGKILL on deadline
  • Interactive confirmation by default
  • Non-root by default; explicit elevation is required when needed

Best Practices

  • Keep allowlist minimal and review regularly
  • Avoid unattended -approve in production
  • Send logs to persistent storage if required for audit

There aren’t any published security advisories