The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code and configuration files from a Spring Boot project to perform the migration.
Ingestion points: The agent reads Java source files, XML/Gradle build descriptors, and configuration files (application.properties, .yaml) from the user-provided source directory into its context.
Boundary markers: There are no explicit delimiters or "ignore embedded instructions" warnings provided to the agent to prevent it from accidentally executing malicious commands that could be hidden in code comments or string literals within the project being migrated.
Capability inventory: The skill grants the agent the ability to execute shell commands (Maven, Gradle, grep), perform extensive file system operations (read/write), and make network requests to external APIs.
Sanitization: The skill does not implement sanitization or filtering of the source code content before it is processed by the AI or used in command-line operations.
[COMMAND_EXECUTION]: The skill frequently executes build and analysis commands on the project code as part of its primary migration logic.
Evidence: Found in SKILL.md (Execution Protocol step 5) and multiple modules (e.g., modules/code/code.md, modules/frontend/frontend.md). It specifically runs ./mvnw clean compile or ./gradlew clean compileJava after each migration phase to verify the build.
Risk: While necessary for the migration task, a malicious project could contain build-time hooks (such as custom Maven plugins or Gradle tasks) that execute arbitrary code when the agent attempts to compile the project. This represents an inherent risk when the agent is used on projects of unknown origin.

migrate-spring-to-quarkus