The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes ./mvnw validate (or mvn validate) and the plantuml CLI as part of its core workflow to ensure project integrity and to render architectural diagrams. These operations are performed within the local environment using standard development tooling.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external and untrusted data, including Java source code, SQL schemas, and repository documentation, to derive architectural patterns. This represents a vulnerability surface where malicious content embedded in the analyzed codebase could attempt to influence the agent's behavior, although the risk is low and inherent to the skill's primary purpose.
[REMOTE_CODE_EXECUTION]: The instructions include specific safety constraints that prohibit the generation of remote !include URLs in PlantUML diagrams and mandate the use of trusted local or vendored libraries, effectively mitigating supply chain risks associated with remote diagram components.

033-architecture-diagrams