The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is explicitly designed to programmatically extract, handle, and cache sensitive Google session cookies, specifically __Secure-1PSID and __Secure-1PSIDTS.
Evidence: Cookie management logic in scripts/gemini-webapi/utils/get-access-token.ts and storage in scripts/gemini-webapi/utils/cookie-file.ts.
[DATA_EXFILTRATION]: The skill employs browser automation techniques to exfiltrate session data from the user's environment. It uses the Chrome DevTools Protocol (CDP) to attach to active browser sessions or read directly from Chrome profile directories to retrieve authentication tokens.
Evidence: implementation in scripts/gemini-webapi/utils/load-browser-cookies.ts utilizing the baoyu-chrome-cdp library to target gemini.google.com cookies.
[COMMAND_EXECUTION]: The skill executes shell commands to perform environment discovery and path resolution, particularly on Windows and WSL systems.
Evidence: scripts/gemini-webapi/utils/paths.ts uses execSync to run cmd.exe /C "echo %USERPROFILE%" and wslpath.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it interpolates untrusted data from local files and user input directly into LLM prompts without sanitization or boundary markers.
Ingestion points: User-provided text and content from local files via the --promptfiles argument in scripts/main.ts.
Boundary markers: Absent; no delimiters are used to separate user data from instructions.
Capability inventory: The skill can read local files, execute environment discovery commands, and perform network requests to Google APIs.
Sanitization: No evidence of escaping or filtering of interpolated content was found.

baoyu-danger-gemini-web