subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external implementation plans and passes it directly to subagents for execution.\n
- Ingestion points: Implementation plan files (e.g.,
docs/plans/feature-plan.md) referenced inSKILL.md.\n - Boundary markers: The templates in
implementer-prompt.mdandspec-reviewer-prompt.mdlack explicit delimiters or instructions to the subagent to ignore embedded commands within the task text.\n - Capability inventory: The implementer subagent is granted file system access to implement code and the ability to commit changes to git (
SKILL.md,implementer-prompt.md).\n - Sanitization: The skill does not perform automated sanitization of the input text; however, the workflow includes a mandatory two-stage review process (spec compliance and code quality) which acts as a manual/agent-based validation gate.
Audit Metadata