skills/pytorch/executorch/qualcomm/Gen Agent Trust Hub

qualcomm

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • System Binary Management: The buck_parity.md file includes steps to download the buck2 build tool from an external repository and install it to a system path using elevated privileges. This is a common requirement for configuring the build environment.
  • Automated Refactoring Loop: The skill utilizes an iterative process to identify and resolve build errors by executing build commands and applying predefined code recipes. This automation helps maintain consistency across different build systems.
  • Dynamic Tool Generation: The qnn_intermediate_debugger.md file guides the agent in creating custom debugging scripts by analyzing and augmenting user-supplied script files. This enables fine-grained accuracy analysis for QNN models.
  • External Dependency Retrieval: The skill automates the acquisition of necessary SDKs and build tools (QNN SDK, Android NDK, and buck2) from official sources, ensuring the environment is correctly provisioned for building and testing.
  • Indirect Prompt Injection Consideration: The script generation workflow represents an indirect prompt injection surface as it involves processing user-provided code. (1) Ingestion points: Path to user-provided scripts in qnn_intermediate_debugger.md. (2) Boundary markers: None explicitly defined for script content ingestion. (3) Capability inventory: Includes file writing, shell command execution (buck2, build.sh), and device interaction via adb. (4) Sanitization: Relies on developer verification of model outputs and structures to maintain debugging integrity.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 03:10 PM
Security Audit — agent-trust-hub — qualcomm