The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary function is to execute code and manage environments on a remote platform (Upstash Box) via a central MCP server (mcp.upstash.com).
[COMMAND_EXECUTION]: Provides the box_exec tool, which allows for the execution of arbitrary shell commands within the remote Linux container.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted data from external sources and has significant capabilities.
Ingestion points: Reads repository content via box_git clone and processes web page content via box_browser (screenshots and live view).
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands in the ingested data.
Capability inventory: Includes arbitrary shell execution (box_exec), Git write access (box_git push/create_pr), and external data exfiltration capabilities via Upstash Blob storage.
Sanitization: No specific sanitization or filtering of external content is described.
[PRIVILEGE_ESCALATION]: The documentation notes that the remote environment allows passwordless sudo access for tasks like package installation.
[EXTERNAL_DOWNLOADS]: Suggests downloading and executing external tools such as the ttyd static binary from GitHub and installing packages via apt-get or npm within the sandbox.
[DYNAMIC_EXECUTION]: Encourages the dynamic creation and execution of shell scripts and commands to perform tasks like environment setup and file editing within the container.

upstash-box-remote-work