subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates a multi-agent workflow where a controller agent ingests external development plans and implementation reports, creating a surface for indirect prompt injection.
- Ingestion points: The workflow involves reading external plan files (e.g.,
docs/plans/feature-plan.md) and processing implementation reports from subagents as defined inSKILL.mdandtemplates/implementer.md. - Boundary markers: The prompt templates utilize markdown headers (e.g.,
## Task Description,## What Was Implemented) to delimit external content, which provides structure but does not fully isolate the agent from potentially malicious instructions embedded in the plans. - Capability inventory: The skill involves executing shell commands for testing, linting, and type-checking, as well as file modifications and git operations (commit, diff) across all templates.
- Sanitization: The skill lacks explicit automated sanitization for external content, instead relying on a two-stage review process (spec compliance and code quality) to catch deviations or harmful implementation details.
Audit Metadata