The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md interpolate user-provided links directly into a shell command template: python3 .../parse-douyin.py <抖音链接>. This allows an attacker to execute arbitrary shell commands by including shell metacharacters (such as semicolons or backticks) in the input.
[EXTERNAL_DOWNLOADS]: The parse-douyin.py script makes network requests to external domains (iesdouyin.com and CDN servers) and downloads media files to the local /tmp directory.
[PROMPT_INJECTION]: The skill parses HTML from external Douyin URLs and extracts video descriptions that are displayed by the agent. Ingestion points: External URLs fetched via the requests library in parse-douyin.py. Boundary markers: No delimiters or instructions are used to prevent the agent from obeying instructions embedded in the external content. Capability inventory: The skill environment allows shell execution, network access, and file writing. Sanitization: Extracted metadata is sanitized for filename safety using regex but is not sanitized for potential prompt injection payloads that could manipulate the agent.

douyin-download