The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill contains standard shell commands for the LLVM toolchain, including compilation (clang), optimization (opt), and assembly generation (llc). These commands are necessary for the skill's intended purpose and do not use elevated privileges.
[DATA_EXPOSURE & EXFILTRATION] (SAFE): The analysis found no evidence of sensitive data access or exfiltration. The skill operates on local source files provided by the user.
[REMOTE_CODE_EXECUTION] (SAFE): No patterns of remote code execution or downloading of external scripts were identified.
[INDIRECT PROMPT INJECTION] (LOW): This skill defines an attack surface by processing untrusted data (C source code and LLVM IR) through compiler utilities.
Ingestion points: Files like src.c, src.ll, and src.bc are read and processed.
Boundary markers: None identified.
Capability inventory: Subprocess execution of clang, opt, llc, and llvm-dis in SKILL.md.
Sanitization: None identified. Maliciously crafted source files could potentially exploit vulnerabilities in the underlying toolchain, though this is a standard risk for development-related skills.

llvm