The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary purpose is to execute code within isolated environments using the mcp__flow-nexus__sandbox_execute tool. This tool allows for running JavaScript, Python, and other languages inside E2B sandboxes.
[DYNAMIC_EXECUTION]: The skill facilitates the creation and execution of dynamic code blocks at runtime. This is a core feature of a development sandbox and is managed through structured tools like mcp__flow-nexus__sandbox_create and mcp__flow-nexus__sandbox_execute.
[INDIRECT_PROMPT_INJECTION]: The skill handles potentially untrusted data through code execution and file uploads, creating a vulnerability surface for indirect injection.
Ingestion points: The code parameter in mcp__flow-nexus__sandbox_execute and the content parameter in mcp__flow-nexus__sandbox_upload are primary entry points for external data.
Boundary markers: The instructions do not define specific delimiters for untrusted data, though they emphasize the use of isolated environments.
Capability inventory: The skill provides capabilities for sandbox creation, code execution, file uploads/downloads, and environment management.
Sanitization: The skill instructions recommend quality standards such as "Secure environment variable management" and "security isolation" to mitigate risks associated with processing untrusted content.

agent-sandbox