bmad-quick-dev-new-preview

Pass

Audited by Gen Agent Trust Hub on Mar 17, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted user input ('intent') to drive automated code generation, which is a common surface for indirect prompt injection.
  • Ingestion points: User input is ingested and clarified in steps/step-01-clarify-and-route.md.
  • Boundary markers: The workflow utilizes <frozen-after-approval> tags within the tech-spec-template.md to clearly delimit human-vetted intent from agent-generated plans.
  • Capability inventory: The skill has the capability to modify the local filesystem ('implement the intent') as described in steps/step-03-implement.md.
  • Sanitization: The skill implements a robust defense-in-depth review process in steps/step-04-review.md. It invokes specialized sub-agents ('Blind hunter', 'Edge case hunter') without conversation context to perform adversarial analysis on generated code, specifically designed to detect malicious or unintended logic before completion.
  • [COMMAND_EXECUTION]: The skill is authorized to perform filesystem operations to implement user-requested code changes.
  • Implementation: In steps/step-03-implement.md, the agent is instructed to implement code artifacts based on a technical specification.
  • Safety Controls: The instructions contain strict operational constraints, including 'No push. No remote ops' during implementation and a 'NEVER auto-push' rule in the final presentation step (steps/step-05-present.md), ensuring that all changes remain local and subject to human approval.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 17, 2026, 07:00 AM
Security Audit — agent-trust-hub — bmad-quick-dev-new-preview