building-architecture-models
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions provide MATLAB code snippets that use the
deletefunction to remove project artifact files like.xmlprofiles and.slmxlink files to facilitate clean model rebuilds. These operations are limited to specific, expected filenames within the project structure.\n- [INDIRECT_PROMPT_INJECTION]: The skill implements a data ingestion surface where external requirement descriptions are parsed for use in quantitative analysis.\n - Ingestion points: System requirement descriptions which may be imported from external sources like Word or Excel, as noted in
references/flp-reference-workflow.md.\n - Boundary markers: None present in the prompt templates used for data processing.\n
- Capability inventory: The skill utilizes
evaluate_matlab_codeto interact with System Composer and Requirements Toolbox APIs for model creation and analysis.\n - Sanitization: The
parseBudgetValuefunction inreferences/system-composer-api.mduses regular expressions to extract specific numeric tokens, providing validation that prevents raw text from being interpreted as code.\n- [SAFE]: All external references point to official MathWorks domains, and the skill follows established engineering practices for the MATLAB environment.
Audit Metadata