The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute build commands such as ./mvnw compile and mvn clean verify. This involves executing scripts contained within the target project's directory, which is a standard but sensitive operation in development workflows.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, review, and refactor existing project source code and configuration files. This creates a vulnerability surface where maliciously crafted content in the analyzed project could attempt to influence the agent's behavior.
Ingestion points: The agent is instructed to read the current project setup and configuration files (e.g., YAML, Java source files) in the first step of the workflow.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings to separate project data from the agent's operational logic.
Capability inventory: The agent has the capability to write to the file system (Step 3: Apply changes) and execute shell commands (Step 4: Run verification).
Sanitization: No explicit sanitization, validation, or filtering of the project content is specified before processing or applying changes.

504-frameworks-micronaut-security