subagent-driven-development

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFE
Full Analysis
  • Subagent Workflow Management: The skill orchestrates the execution of software implementation plans by dispatching specialized subagents for distinct tasks. This approach uses context isolation and two-stage reviews (spec compliance and code quality) to maintain high output standards.
  • Indirect Prompt Injection Surface: The skill processes external implementation plans, which presents a surface for indirect prompt injection if the input files contain instructions designed to manipulate the subagents.
  • Ingestion points: Task text and context are extracted from external plan files (e.g., in docs/superpowers/plans/) and used to populate subagent prompts in implementer-prompt.md and spec-reviewer-prompt.md.
  • Boundary markers: The prompt templates utilize Markdown headers to delimit task descriptions from the agent's core instructions, providing basic logical separation.
  • Capability inventory: Tasks involve implementation, testing, and version control operations within the development environment.
  • Sanitization: Content from the plan files is interpolated into prompts without automated escaping or filtering.
  • Independent Review Mechanisms: The 'Spec Compliance Reviewer' is explicitly instructed to verify code directly rather than relying on the implementer's status report. This design provides a verification layer against potentially inaccurate or incomplete reporting by the implementation subagent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 02:17 PM
Security Audit — agent-trust-hub — subagent-driven-development