The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill is designed specifically to manage remote code execution workflows using the E2B sandbox service. It provides tools to create environments, execute code, and manage file lifecycles.
[COMMAND_EXECUTION]: The tool mcp__flow-nexus__sandbox_execute allows the agent to run arbitrary code in multiple languages (Node.js, Python, etc.) within the created sandboxes.
[EXTERNAL_DOWNLOADS]: The skill supports the dynamic installation of packages from external registries (e.g., npm, pip) as part of the sandbox configuration process, specifically via the install_packages parameter.
[PROMPT_INJECTION]: The skill manages a significant attack surface for indirect prompt injection because it ingests and executes data from untrusted sources.
Ingestion points: Data enters the agent's context through the code parameter in execution calls and the content parameter in file upload calls.
Boundary markers: The provided instructions do not include specific delimiters or warnings to ignore instructions embedded within the code or files being processed.
Capability inventory: The skill possesses the ability to execute code, install software, and perform network operations within the isolated sandbox environment.
Sanitization: There is no evidence of input validation or sanitization for the code snippets or file contents handled by the tools.

agent-sandbox