ruby-cext-memory-truffle-hunt

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The static analysis tools in the references/ folder (sweep_unmarked.py, sweep_escaped_conversion.py, sweep_static_values.py) are designed to ingest and parse untrusted C/C++ source code. This represents a functional attack surface common to security auditing software. * Ingestion points: Analysis scripts read source files from user-specified directories using the pathlib module. * Boundary markers: Absent; the tools parse code directly to identify vulnerability patterns. * Capability inventory: Includes directory management via subprocess.run and low-level memory access in the Ruby harness. * Sanitization: The scripts use complex regular expressions to process source code, which is a potential target for adversarial input designed to mislead the parser.
  • [COMMAND_EXECUTION]: System commands are used to support the auditing and verification processes. * sweep_unmarked.py invokes the system cp command using subprocess.run to organize test fixtures for its self-check routine. * references/repro-class-b-psych.rb executes shasum through shell backticks to confirm the integrity of loaded binary extensions. * references/harness.rb uses fork and process waiting to safely handle memory inspection in child processes. These actions are documented and integrated into the skill's primary function.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 06:19 AM
Security Audit — agent-trust-hub — ruby-cext-memory-truffle-hunt