bmad-build
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local Python scripts, specifically
render_skill.pyandtickets.py, from the project's_bmaddirectory using theuvtool. These executions are central to the skill's workflow management. - [INDIRECT_PROMPT_INJECTION]: The skill reviews user-provided intent and generated code diffs using context-free subagents. While this ingestion of untrusted data represents an attack surface, the subagents are constrained by instructions to perform "inspection only," and specific boundary markers are used in prompts to isolate data from instructions.
- [DYNAMIC_EXECUTION]: The workflow steps are dynamically generated using a templating system, allowing the agent to adapt its instructions based on the resolved route (e.g., oneshot vs. full) and current progress.
Audit Metadata