The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/sogou_wechat.py uses subprocess.run to execute scp and ssh commands. These commands are used to transfer a dynamically generated Python script to a remote host and execute it, which constitutes a high-risk remote code execution pattern.
[PRIVILEGE_ESCALATION]: In scripts/sogou_wechat.py and scripts/fetch_china.py, the skill attempts to write command strings to /root/router-agent/cmd-queue. Interacting with the /root directory is a sensitive operation that requires superuser privileges and is a strong indicator of privilege escalation or an assumption of excessive permissions.
[DYNAMIC_EXECUTION]: The function sogou_wechat_search_via_ssh in scripts/sogou_wechat.py dynamically constructs a Python script as a string and executes it on a remote machine. Constructing executable code from string templates is a dangerous pattern that can lead to code injection.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to fetch content from external, untrusted social media platforms (X, WeChat, Weibo, etc.).
Ingestion points: Data is fetched in scripts/fetch_tweet.py and scripts/fetch_china.py.
Boundary markers: The skill does not implement delimiters or 'ignore' instructions for the fetched content before returning it to the agent.
Capability inventory: The skill possesses file-write and subprocess capabilities, which increases the potential impact of an injection attack.
Sanitization: There is no specific sanitization to prevent malicious instructions embedded in tweets or articles from influencing the AI agent's behavior.

x-tweet-fetcher