The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md construct a shell command string using user-provided input (keywords). If the keyword contains shell metacharacters such as semicolons or backticks, it could lead to arbitrary command execution on the host system.
[EXTERNAL_DOWNLOADS]: The skill environment setup involves downloading and installing the DrissionPage package from the Python Package Index (PyPI).
[REMOTE_CODE_EXECUTION]: The skill executes a custom local Python script (search.py) which performs network operations and browser automation to fetch data from external servers.
[PROMPT_INJECTION]: Indirect Prompt Injection Risk: The skill processes video titles and author names from Douyin's search results without sanitization. This external data is subsequently analyzed by the agent, creating a vector where malicious content in search results could influence the agent's behavior.
Ingestion points: scripts/search.py listens for and parses network traffic from douyin.com search APIs.
Boundary markers: None identified in the prompt templates or data processing logic.
Capability inventory: Shell execution capabilities for environment setup and script execution.
Sanitization: None; the agent processes the raw JSON output from the search script.

laohan-douyinsousuo