The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, including existing Simulink models (via model_read) and web search results (via web_search). This ingestion of untrusted data presents a theoretical surface for indirect prompt injection if malicious instructions were present in model metadata or external documentation. However, the workflow relies on structured markdown templates and defined review gates that require the agent to verify technical details before finalizing the specification.
[DYNAMIC_EXECUTION]: The workflow employs evaluate_matlab_code to verify API signatures and block behavior during architecture design. This dynamic execution is targeted at technical verification within the MATLAB environment to ensure that specifications match the intended implementation logic and is consistent with the skill's primary purpose.

specifying-mbd-algorithms