baoyu-post-to-weibo

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of system commands to automate browser processes and manage the clipboard. It invokes powershell.exe on Windows, osascript and swift on macOS, and xclip, wl-copy, xdotool, or ydotool on Linux. Evidence is found in scripts/copy-to-clipboard.ts and scripts/paste-from-clipboard.ts for cross-platform support.
  • [DYNAMIC_EXECUTION]: To handle rich text and images in the clipboard, the skill dynamically constructs and executes Swift source code on macOS and PowerShell commands on Windows. Evidence includes getMacSwiftClipboardSource and copyImageWindows in scripts/copy-to-clipboard.ts.
  • [EXTERNAL_DOWNLOADS]: The skill depends on third-party packages baoyu-chrome-cdp and baoyu-md for its core operations, and its execution instructions involve using npx -y bun to provide the runtime environment. Evidence is in scripts/package.json and SKILL.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests arbitrary text and Markdown files provided by the user to generate social media posts.
    1. Ingestion points: Markdown and plain text arguments in weibo-post.ts and weibo-article.ts.
    1. Boundary markers: None identified in the input processing scripts.
    1. Capability inventory: Significant capabilities including system command execution (spawnSync) and browser script execution (Runtime.evaluate).
    1. Sanitization: Uses JSON.stringify to escape content before inserting it into the browser context, which mitigates direct script injection into the Weibo editor.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:47 AM
Security Audit — agent-trust-hub — baoyu-post-to-weibo