The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to invoke Java and the PlantUML library. These commands are well-defined and follow standard CLI usage patterns for the tool.
Evidence in SKILL.md: Commands like java -jar "./assets/plantuml.jar" --format svg "<input_file>" are used for rendering.
Evidence in SKILL.md: The skill includes a 'Quick Check' phase using java -version to ensure environment readiness.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive file paths are accessed, and no network operations to non-whitelisted domains are initiated. The skill operates entirely on local inputs provided by the user.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. It relies on a local JAR file (./assets/plantuml.jar) and standard system Java.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-provided PlantUML text, it does so by passing it to an external rendering engine (PlantUML) which does not interpret the content as instructions for the AI agent. Boundary markers like @startuml and @enduml are used to isolate the data.

plantuml-renderer