java-upgrade
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface because it is designed to read and process external Java source code and configuration files (such as pom.xml and build.gradle).
- Ingestion points: The Phase 0 assessment and all six transformation groups require the agent to analyze local application code.
- Boundary markers: The skill does not explicitly define delimiters to separate its instructions from the content of the files being analyzed.
- Capability inventory: The skill utilizes subprocess calls for build automation (mvn, gradle), code diagnostics (jdeps, jdeprscan), and containerization (docker).
- Sanitization: There is no specific requirement for the agent to sanitize or validate the source code content before performing modifications. This risk is considered low as it is intrinsic to the skill's primary function of code modernization.
- [COMMAND_EXECUTION]: The skill requires the execution of multiple command-line tools to achieve its objectives.
- Evidence: The instructions mandate the use of Maven/Gradle plugins, OpenRewrite recipes, and Docker build commands. These actions are justified by the skill's primary purpose and follow documented safe practices, such as the use of Amazon Corretto base images and non-root execution contexts.
Audit Metadata