Skills
Navigate
skills/fabriqaai/codex-codebank/codex-banked-resets/Gen Agent Trust Hub

codex-banked-resets

Pass

Audited by Gen Agent Trust Hub on Oct 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The script uses node:child_process.spawn to open the user's default web browser during the OAuth login flow.
  • Evidence: spawn(command, args, { detached: true, stdio: "ignore" }) in scripts/codex-banked-resets.mjs triggers open (macOS), cmd (Windows), or xdg-open (Linux) with a login URL.
  • [EXTERNAL_DOWNLOADS]: The skill makes network requests to official OpenAI and ChatGPT API endpoints to fetch profile and credit data.
  • Evidence: scripts/codex-banked-resets.mjs targets auth.openai.com and chatgpt.com/backend-api for authentication and data retrieval.
  • [DATA_EXPOSURE]: The skill accesses local configuration files to retrieve existing authentication tokens.
  • Evidence: The script reads ~/.codex/auth.json and files in ~/.codex/accounts/ to deduplicate and use stored access tokens.
  • [SAFE]: The skill follows security best practices for handling sensitive data, such as explicitly instructing the agent and script to avoid printing raw tokens, cookies, or internal IDs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 3, 2026, 12:45 PM