java-architect

Warn

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill workflow directs the agent to execute shell commands using Maven and Gradle wrappers (e.g., ./mvnw verify, ./gradlew check). This capability presents a risk if the agent is operating on a project with malicious wrapper scripts, as it could lead to arbitrary code execution in the local environment.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It is designed to read and analyze external project content, such as source code, configuration files, and domain models, to provide architectural guidance. Maliciously crafted content within these analyzed files could be used to influence the agent's behavior.\n
  • Ingestion points: Workflow steps in SKILL.md requiring the agent to review project structure and domain boundaries.\n
  • Boundary markers: Not implemented; the skill lacks explicit markers to distinguish between its own instructions and the user data being processed.\n
  • Capability inventory: Access to shell commands via build tools as defined in SKILL.md.\n
  • Sanitization: Not implemented; the skill does not specify validation or filtering for ingested project data.\n- [PROMPT_INJECTION]: The skill metadata contains deceptive information. The author field in SKILL.md lists https://github.com/Jeffallan, which contradicts the verified author context of neg0diay. Misleading metadata regarding a skill's origin is a security concern for provenance verification.\n- [EXTERNAL_DOWNLOADS]: The configuration in references/spring-boot-setup.md references the Maven build system, which fetches dependencies from the well-known Maven Central repository. This is standard and expected behavior for the specified development environment.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 3, 2026, 09:32 AM
Security Audit — agent-trust-hub — java-architect