The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from Simulink models to generate requirement summaries, which could be exploited if a model contains malicious instructions intended to influence the agent's output.
Ingestion points: Model data is retrieved via model_read, model_overview, model_query_params, and model_resolve_params tools as described in the SKILL.md workflow.
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the model metadata.
Capability inventory: The agent can execute MATLAB commands via evaluate_matlab_code and perform file system writes for .slreqx and .yaml artifacts.
Sanitization: No explicit sanitization is described, although the skill mandates that all generated content be marked as 'DRAFT' and requires human review before use.
[DYNAMIC_EXECUTION]: The skill dynamically generates and executes MATLAB script snippets to perform requirements management tasks.
Evidence: The 'Path A' workflow in SKILL.md constructs MATLAB code strings (e.g., using slreq.new, add, and slreq.createLink) and executes them through the evaluate_matlab_code tool.

generate-requirement-drafts