subagent-driven-development

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill architecture promotes security through isolation and mandatory quality gates. Each task is executed in a fresh context, and all code is subjected to independent spec compliance and quality reviews.
  • [PROMPT_INJECTION]: The workflow manages the risk of indirect prompt injection from implementation plans through deliberate prompt construction and validation loops.
  • Ingestion points: Implementation plans (e.g., docs/superpowers/plans/feature-plan.md) are parsed to extract task descriptions.
  • Boundary markers: Subagent prompt templates use explicit markdown headers and instructions to delimit injected plan content.
  • Capability inventory: Subagents are authorized to modify files and execute tests within the project scope.
  • Sanitization: The process relies on a two-stage review loop (Spec Reviewer and Code Quality Reviewer) to verify that implementers only performed requested actions and that the code adheres to safety and quality standards.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 11:14 AM