The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation describes interactions with well-known and trusted external services such as api.openai.com for LLM calls and pypi.org for package management. It also demonstrates usage of Amazon S3 for storing snapshot artifacts.- [COMMAND_EXECUTION]: Implementation examples for sandbox management use the Python subprocess module to execute system commands for hypervisor control (e.g., runsc and firecracker). These are foundational to the skill's purpose of managing isolated execution boundaries.- [REMOTE_CODE_EXECUTION]: The skill is architected to facilitate the execution of code within isolated microVMs or containers. It provides detailed guidance on maintaining strict isolation through seccomp filters and hardware-level virtualization to prevent host compromise.- [SAFE]: No malicious prompt injection, data exfiltration patterns, or unauthorized credential usage were detected. Although several reference files contain highly repetitive and filler content, no malicious payloads or hidden instructions were found within these sections.

sandbox-execution