subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill coordinates subagents to execute implementation plans, which involves ingesting untrusted task data from external files.
- Ingestion points: Task descriptions and requirements are read from implementation plan files (e.g., docs/superpowers/plans/feature-plan.md) and passed into prompts within implementer-prompt.md and spec-reviewer-prompt.md.
- Boundary markers: The templates use Markdown headers to separate task text from instructions, but they lack explicit directives for the subagents to ignore potential instructions embedded within the untrusted task text.
- Capability inventory: Subagents (using general-purpose tools and superpowers:test-driven-development) are explicitly instructed to modify code, create new files, and execute tests as part of their tasks, as documented in implementer-prompt.md.
- Sanitization: No sanitization, validation, or escaping of the task requirements is performed before they are interpolated into the subagent's execution context.
Audit Metadata