pyjail

This skill is internally consistent with its stated purpose, but that purpose is offensive: it equips an AI agent to bypass Python sandboxes and execute commands or read files. There is little evidence of credential theft or external exfiltration, but the exploit-focused behavior makes it high risk as an agent skill.

This is an explicit collection of Python jail-escape techniques. It can enable unauthorized file reads and command execution when run in a vulnerable restricted interpreter, but the fragment itself is instructional text and does not demonstrate package malware or execute autonomously.

This is explicit pyjail escape guidance with examples capable of arbitrary code execution and local file disclosure if run in a vulnerable or permissive environment. The fragment itself is documentation, not an active malware payload; its risk is contextual and depends on use in a sandbox-escape setting.