neuron-nki-profiling

Pass

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/identify-neffs.py uses subprocess.run to execute the standard strings utility for inspecting Neuron Executable (NEFF) binary files. The execution is handled safely without shell interpolation.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from NEFF and HLO module files (ingestion points) to identify kernel names and metadata. Capability inventory includes file system access and execution of the strings tool. Boundary markers are absent for these binary formats, and sanitization is performed via JSON decoding of embedded metadata.
  • [DYNAMIC_EXECUTION]: The skill uses the NKI Just-In-Time (JIT) compiler to build hardware-specific kernel binaries at runtime, which is a core and expected feature of the AWS Neuron SDK for performance optimization.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 16, 2026, 11:14 AM