subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from implementation plans and source code, creating a surface for indirect prompt injection that could influence subagent behavior.
- Ingestion points: Task descriptions are extracted from plan files (e.g.,
docs/superpowers/plans/feature-plan.md) and passed to implementer subagents. Source code implementation is read and analyzed by the spec and code quality reviewer subagents. - Boundary markers: The prompt templates (
implementer-prompt.md,spec-reviewer-prompt.md) use Markdown headers to delineate task descriptions and requirements, though these provide only weak separation from the model's instructions. - Capability inventory: The skill facilitates subagents performing implementation, testing, and Git commits, providing a pathway for malicious code to be introduced if a subagent is successfully injected.
- Sanitization: The skill does not implement sanitization or validation of the text extracted from implementation plans or source code before interpolation into subagent prompts.
Audit Metadata