clang

Pass

Audited by Gen Agent Trust Hub on Feb 20, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (LOW): The skill operates by reading and analyzing user-provided source code files, which creates a potential surface for indirect prompt injection via embedded comments.
  • Ingestion points: External C/C++ source files (e.g., src.c, src.cpp) processed by the agent.
  • Boundary markers: None specified; the instructions do not include specific delimiters or 'ignore' warnings for the data being compiled or analyzed.
  • Capability inventory: The skill uses subprocess calls to execute clang, clang-tidy, and llvm-profdata.
  • Sanitization: No sanitization or escaping of file content is described in the workflow.
  • Dynamic Execution (SAFE): The skill provides instructions for the compilation and execution of binary files (e.g., clang ... -o prog_inst and ./prog_inst). This behavior is the primary and expected function of a compiler tool and does not represent a security violation in this context.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 20, 2026, 03:23 PM