The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to manage the browser lifecycle and publication pipeline. \n
scripts/chrome_launcher.py uses subprocess.Popen to start Google Chrome with remote debugging enabled.\n
scripts/chrome_launcher.py uses subprocess.run to execute netstat and taskkill on Windows for process management.\n
The SKILL.md instructions guide the agent to execute various Python scripts (publish_pipeline.py, cdp_publish.py) with user-influenced arguments.\n- [EXTERNAL_DOWNLOADS]: The scripts/image_downloader.py script downloads image files from arbitrary URLs provided by the user or found on webpages using the requests library. These files are stored in a temporary local directory before being uploaded to the social media platform.\n- [DYNAMIC_EXECUTION]: The scripts/cdp_publish.py script interacts with the Chrome browser by executing JavaScript snippets via the Runtime.evaluate method of the Chrome DevTools Protocol. This allows the skill to manipulate the webpage DOM, fill forms, and click buttons programmatically. The skill mitigates risks by using json.dumps to escape content before interpolation.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external URLs, creating a surface for indirect prompt injection.\n
Ingestion points: SKILL.md (Step 2: URL 提取模式) instructs the agent to use WebFetch to extract content and image URLs from a user-provided webpage.\n
Boundary markers: None identified; instructions do not specify the use of delimiters or warnings to ignore instructions within the fetched content.\n
Capability inventory: The skill has capabilities to execute shell commands (subprocess), write and delete files (shutil, os), and control a web browser instance (websockets/CDP).\n
Sanitization: The skill instructions rely on the agent to "summarize" content but do not provide explicit sanitization or filtering logic for the extracted text.

post-to-xhs