The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill targets sensitive configuration files at ~/.cursor/mcp.json and ~/.claude.json, which are standard locations for storing plain-text API keys and Model Context Protocol (MCP) server settings.
[DATA_EXFILTRATION]: Troubleshooting scripts in the skill (specifically in references/troubleshooting.md) extract and print full server configurations, including raw API keys from environment variables, directly into the agent's output history.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via mcporter and python3 heredocs to interact with the local filesystem and invoke external tool functions.
[EXTERNAL_DOWNLOADS]: Setup instructions recommend global installation of the mcporter utility via npm and runtime execution of douyin-mcp-server from external sources.
[REMOTE_CODE_EXECUTION]: The skill uses uvx to dynamically download and execute the douyin-mcp-server package, which constitutes execution of external code from a remote registry in the shell environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcription data from external videos without proper isolation.
Ingestion points: Video text enters the context via the douyin-mcp.extract_douyin_text tool output.
Boundary markers: No delimiters or 'ignore' instructions are used to separate transcription data from the agent's logic.
Capability inventory: The agent has access to browser tools and mcporter shell execution.
Sanitization: There is no evidence of validation or sanitization of the extracted video text before the agent summarizes it.

douyin-video-analyst