706-technologies-containers-docker
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses standard Maven build commands (
./mvnw compile,mvn clean verify) to ensure the integrity of the Java project before and after containerization changes. These operations are local to the development environment and align with the skill's stated purpose of providing reliable Docker configurations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project data (Dockerfiles, Maven settings, CI jobs) as specified in the workflow. While this constitutes an attack surface, the risk is inherent to code-review agents and is mitigated by the skill's specific focus on security best practices and the requirement for manual verification of build outputs. Evidence chain: 1. Ingestion points: local Dockerfiles and build scripts (SKILL.md). 2. Boundary markers: absent. 3. Capability inventory: Maven execution and file system writes. 4. Sanitization: absent.
- [SAFE]: The skill explicitly promotes secure defaults, such as avoiding secrets in image layers, implementing non-root user execution, and using minimal base images like Alpaquita. These instructions significantly improve the security posture of the resulting container images.
Audit Metadata