codex-auth-helper-chrome-extension

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs extracting session cookies/access tokens and placing them verbatim into a downloadable auth.json (and potentially the UI), which requires the agent to read and output secret credentials directly.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). High risk — the links include a typosquat domain (chatgpt.com instead of the official chat.openai.com), an unfamiliar/site (ara.so), and a small/unknown GitHub repo offering a Chrome extension that reads session cookies and exports tokens, which are classic indicators for credential-theft or malware distribution.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The extension deliberately reads ChatGPT session cookies and generates/saves an auth.json containing access/id tokens (with options for automatic or batch export), which is an explicit credential-extraction capability that can be abused for data exfiltration.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The skill’s runtime workflow reads ChatGPT session data via fetch('https://chatgpt.com/api/auth/session') and cookie extraction from chatgpt.com; this is outsider-authored free text (ChatGPT’s API responses / page-originated content) being ingested into the extension’s JS/LLM context.