subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill defines a process for reading implementation plans from local files and passing the contents to subagents as instructions, creating an indirect prompt injection surface.
- Ingestion points: The skill instructions direct the agent to read plan files (e.g.,
docs/plans/feature-plan.md) using theread_filetool (SKILL.md). - Boundary markers: None identified. The instructions specify providing the "full task text directly in context" to subagents without delimiters or instructions to ignore embedded commands.
- Capability inventory: The subagents are dispatched via
delegate_taskwith access to theterminalandfiletoolsets, enabling them to execute arbitrary commands, run tests, and modify the source code (SKILL.md). - Sanitization: None. The content of the plan files is treated as authoritative context for the subagent's goal.
Audit Metadata