The Agent Skills Directory

[COMMAND_EXECUTION]: Phase 7.1 instructs the agent to run the project's build command to validate layout changes. This is a common and expected operation for developer-centric tools to ensure code correctness after automated modifications.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external data. 1) Ingestion points: The skill reads codebase style files and $ARGUMENTS. 2) Boundary markers: No specific delimiters or safety warnings for external content are defined. 3) Capability inventory: The skill can modify files and execute project build commands. 4) Sanitization: No sanitization of ingested content is mentioned. This remains a low-risk finding typical of tools that analyze user-provided codebases.
[SAFE]: The skill focuses entirely on UI/UX adaptation using standard framework features (CSS Container Queries, Flutter LayoutBuilder). No evidence of credential harvesting, remote code execution from unknown sources, or persistence mechanisms was found.

design-adapt