The Agent Skills Directory

[SAFE]: The skill implements a strict 'Extract only, Never execute' policy, ensuring that any extracted binaries or scripts are treated as data for inspection rather than executable code.
[SAFE]: Disk images (.dmg, .iso) are mounted with security-conscious flags like -nobrowse and -readonly to prevent the operating system from triggering automatic file previews or execution.
[SAFE]: All shell commands use robust path quoting (e.g., "") to protect against command injection vulnerabilities that could arise from maliciously crafted filenames.
[SAFE]: The extraction workflow includes an explicit pre-check for encryption using 7z l -slt to prevent the agent from hanging indefinitely on hidden password prompts.
[SAFE]: A recursion limit of three levels and a cycle guard are implemented to mitigate the risk of zip bombs and circular directory structures that could lead to resource exhaustion.
[SAFE]: The skill leverages industry-standard tools (7z, unar, innoextract, xar) from established package managers and official registries, which are considered safe sources.

unbox-anything