The Agent Skills Directory

[SAFE]: The skill is designed for legitimate TikTok research and analytics through the SandBase gateway. It includes specific operating principles that restrict the agent to read-only interactions, explicitly forbidding posting, liking, or following on the platform.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as TikTok video descriptions, creator bios, and hashtag metadata through the tiktok_app_v3_* suite of tools. While this constitutes an indirect prompt injection surface, the skill lacks high-risk capabilities—such as arbitrary command execution, file system writes, or sensitive data exfiltration—that would be necessary to exploit this surface effectively.
[COMMAND_EXECUTION]: The skill references sandbase_describe_tool and sandbase_call_tool to interact with the vendor's API. These are standard tool-calling patterns within the platform's ecosystem and are scoped specifically to the provided TikTok research capabilities, representing no risk of arbitrary shell command execution.

tiktok-research