re-java
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches decompilation tools from established software repositories.
- Evidence: Downloads
cfr.jarfrom the tool's official website andprocyon-decompiler.jarfrom the Maven Central repository (SKILL.md). - [REMOTE_CODE_EXECUTION]: Executes downloaded binary JAR files as part of the decompilation workflow.
- Evidence: The instructions involve running
java -jar cfr.jarandjava -jar procyon-decompiler.jaron the local machine after downloading them (SKILL.md). - [COMMAND_EXECUTION]: Utilizes system package managers and command-line utilities for environment setup and file processing.
- Evidence: Includes commands for installing packages via
apt,dnf,pacman,brew, andchoco, as well as usingunzipandjarfor file extraction (SKILL.md). - [INDIRECT_PROMPT_INJECTION]: The skill processes external Java bytecode files, which introduces an indirect prompt injection surface.
- Ingestion points: The skill ingests untrusted
jar,war, andclassfiles for analysis (SKILL.md). - Boundary markers: No explicit boundary markers or instructions to ignore embedded content within the analyzed files are provided to the agent.
- Capability inventory: Includes file system access (
unzip,jar), command execution (javap,java -jar), and dynamic debugging tools (jdb,frida) (SKILL.md). - Sanitization: No sanitization or validation of the contents of the processed bytecode is described.
- [DYNAMIC_EXECUTION]: Employs dynamic analysis tools and runtime instrumentation.
- Evidence: Provides instructions for using the Java Debugger (
jdb) to evaluate code at runtime and utilizingFridato hook Java methods and modify their implementation during execution (SKILL.md).
Audit Metadata