The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions in SKILL.md and automated checks in scripts/check-permissions.ts to install the Bun runtime by piping a remote script directly to the shell (curl -fsSL https://bun.sh/install | bash). This is a high-risk pattern that executes unverified code from an external source.
[COMMAND_EXECUTION]: The skill frequently executes system-level commands to automate browser interaction and bypass security restrictions.
Uses osascript on macOS to programmatically simulate 'Cmd+V' and 'Cmd+C' keystrokes to interact with the WeChat editor.
Utilizes powershell.exe on Windows and xdotool/ydotool on Linux for similar keyboard and clipboard automation.
Launches Google Chrome via spawn with high-privilege flags such as --remote-debugging-port and --disable-blink-features=AutomationControlled to evade automation detection.
Executes lsof and ss commands to scan for listening TCP ports to detect existing Chrome debugger instances.
[EXTERNAL_DOWNLOADS]: The scripts/md-to-wechat.ts file includes functionality to automatically download image files from remote URLs found in user-provided markdown files. These files are stored in temporary directories and subsequently uploaded to WeChat servers without integrity verification or domain restriction.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it ingests untrusted markdown data and parses metadata (title, author, summary) and image URLs to perform automated actions. An attacker could craft a markdown file that, when processed by the skill, triggers unintended network requests or manipulates the content of the final WeChat post without the user's explicit consent.

baoyu-post-to-wechat