bun-ffi
Warn
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documents the use of
dlopenandCFunctionfrom thebun:ffimodule. This allows the agent to load shared libraries (.so,.dylib,.dll) from the file system and execute their functions natively. This is a powerful primitive that bypasses JavaScript safety sandboxes and can be used to execute arbitrary compiled code. - [COMMAND_EXECUTION]: The instructions demonstrate how to bind and call standard C library functions like
printfandgetenv. These functions interact directly with the operating system and process environment, providing a mechanism for system-level operations. - [INDIRECT_PROMPT_INJECTION]: The skill involves ingesting data from external shared libraries (e.g., return values typed as
cstringorptr). - Ingestion points: Data enters the agent's context through return values of bound FFI symbols (e.g.,
symbols.getenv("HOME")orsymbols.hash(...)). - Boundary markers: None. The skill does not describe using delimiters to separate library output from instructions.
- Capability inventory: The skill has access to
Bash,Write, andReadtools, and can perform native memory operations via FFI pointers and buffers. - Sanitization: There is no evidence of sanitization or validation of the data returned from native functions before it is processed by the agent.
Audit Metadata