The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides the standard installation command for the Bun runtime in SKILL.md, which fetches a script from Bun's official website (https://bun.sh/install) and pipes it to a shell for execution. This is a common and legitimate practice for software installation from a well-known service.
[COMMAND_EXECUTION]: The skill documents the Bun Shell ($) and Bun.spawn for executing system commands. The documentation includes security considerations, noting that interpolated variables are auto-escaped to prevent shell injection, while also documenting the { raw: str } option for cases where unescaped execution is explicitly required.
[DYNAMIC_EXECUTION]: In references/advanced.md, the skill documents advanced runtime features such as the Foreign Function Interface (FFI) using dlopen and an inline C compiler (cc). These are legitimate, high-performance features of the Bun runtime intended for native library integration.
[INDIRECT_PROMPT_INJECTION]: The skill documents capabilities for ingesting data from external sources, including HTTP servers (Bun.serve), network requests (fetch), and file I/O. While these provide an attack surface for indirect prompt injection if an agent were to process untrusted data without sanitization, the skill itself does not contain instructions to override safety guidelines or blindly trust external content.

bun