bmad-build-auto

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a Python script located within the project's directory (_bmad/scripts/render_skill.py) using uv run to drive the automation process.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as planning documents and code diffs, which serves as a potential attack surface.\n
  • Ingestion points: Reads PRDs, architecture guides, and epic contexts in step-01-clarify-and-route.md, as well as unified diffs in step-04-review.md.\n
  • Boundary markers: Employs <intent-contract> delimiters in spec-template.md to isolate core intent from other instructions.\n
  • Capability inventory: Capable of executing shell commands, modifying the file system, and invoking subagents to perform code generation and review.\n
  • Sanitization: Content is processed through specific Markdown templates and specialized subagent prompts designed to maintain focus on technical review and implementation tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 10:42 PM
Security Audit — agent-trust-hub — bmad-build-auto