quota-reporter

Fail

Audited by Gen Agent Trust Hub on May 11, 2026

Risk Level: HIGHDATA_EXFILTRATIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DATA_EXFILTRATION]: The skill's primary function is to extract and transmit full authentication snapshots—including sensitive JWTs and OAuth access tokens—to a remote server. This occurs in scripts/quota_reporters.py via the post_auth_pool_entry function, which sends auth JSON data to the configured auth_pool_url.
  • [CREDENTIALS_UNSAFE]: The skill specifically targets and reads highly sensitive credential files, including ~/.codex/auth.json and Claude's .credentials.json. Additionally, on macOS systems, it utilizes the security find-generic-password command in scripts/quota_reporters.py to programmatically extract credentials from the System Keychain.
  • [COMMAND_EXECUTION]: The installer (scripts/install_quota_guard.py) and various maintenance scripts execute numerous shell commands to configure system-level persistence and notifications. This includes the use of launchctl (macOS), crontab (Linux), and New-ScheduledTask via PowerShell (Windows). The Windows implementation explicitly uses -ExecutionPolicy Bypass to ensure its scripts run without restriction.
  • [EXTERNAL_DOWNLOADS]: The fetch_best_auth function in scripts/quota_reporters.py dynamically downloads authentication blobs from a remote API and writes them directly to local credential files. This mechanism allows for the remote replacement of authentication identities used by the agent and other local tools.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 11, 2026, 03:42 AM