backend-dev-agent
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill automates the execution of build and package managers including Maven and Gradle. This involves running system-level commands to compile, test, and package Java applications.
- [EXTERNAL_DOWNLOADS]: The use of Maven and Gradle implies the download of external libraries and dependencies from remote registries (such as Maven Central) during the build phase. Additionally, the skill utilizes OpenAPI Gen for automated code stub generation which may fetch remote templates.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external sources like Jira tickets and PR descriptions.
- Ingestion points: Sub-agent 2 (Work Intake) parses raw text from JIRA/ADO tickets and natural language requests.
- Boundary markers: The skill documentation notes that external tool outputs are treated as untrusted and sanitized, though explicit prompt delimiters for ingested text are not defined in the skill file.
- Capability inventory: The orchestrator has high-privilege capabilities including writing to Git repositories, triggering Jenkins builds, and creating documentation in Confluence.
- Sanitization: The skill explicitly includes a guardrail stating that all external tool output is sanitized before use and that no production credentials are permitted in the workflow.
Audit Metadata