The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system-level commands to facilitate video downloading and media processing.
Evidence in scripts/jianying_worker.py: Uses subprocess.run(['curl', '-L', '-o', filepath, '-s', '-w', '%{http_code}', mp4_url], ...) to download generated video files from remote URLs.
Evidence in scripts/jianying_worker.py: Utilizes asyncio.create_subprocess_exec to call ffmpeg and ffprobe for video resizing and metadata extraction.
Evidence in scripts/poll_and_download.py: Also uses subprocess.run with curl to fetch video files during polling.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data and user input that could potentially influence the agent's behavior if malicious instructions are embedded.
Ingestion points: The skill ingests a user-provided --prompt via the CLI and scrapes HTML content/metadata from xyq.jianying.com (found in scripts/jianying_worker.py and scripts/find_latest_task.py).
Boundary markers: No explicit boundary markers or "ignore instructions" warnings are present when interpolating user prompts or site content into the automation flow.
Capability inventory: The skill has the ability to execute shell commands, perform network operations via browser automation, and write files to the local system.
Sanitization: There is no evidence of sanitization or filtering for the prompt text or the content scraped from the web interface before it is processed by the agent scripts.

jianying-video-gen