qt-unittest-make
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes local C++ header and source files to extract class structures and function signatures. This creates a surface for indirect prompt injection if the project's source code contains adversarial instructions in comments or string literals designed to manipulate the agent during the test generation process.
- Ingestion points: Source code files are read via the read tool and LSP tools (lsp_document_symbols, lsp_goto_definition) as specified in agent/unittest-generator.md.
- Boundary markers: The instructions do not define specific delimiters for untrusted code content when passed to the sub-agent.
- Capability inventory: The skill possesses write capabilities to generate test files and bash capabilities to compile them.
- Sanitization: No explicit sanitization of extracted source code content is performed before interpolation into prompts.
- [DYNAMIC_EXECUTION]: The skill dynamically generates C++ unit test files from templates and extracted metadata. It then compiles and potentially executes these tests using the system's build tools.
- Evidence: The sub-agent in agent/unittest-generator.md is instructed to generate test code based on templates in resources/templates/ and then run cmake --build to verify the build.
- [COMMAND_EXECUTION]: The agent uses the shell to manage the build environment, configure projects with CMake, and run tests.
- Evidence: agent/unittest-generator.md utilizes the bash tool to execute commands like mkdir, cmake, and ctest for mandatory build verification.
Audit Metadata