subagent-driven-development

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a workflow for executing tasks defined in external plan files, which is vulnerable to indirect prompt injection.
  • Ingestion points: The agent is instructed to read tasks from a [plan-file] which is an external data source entering the agent context.
  • Boundary markers: The provided prompt templates for subagents do not include delimiters or instructions to ignore embedded malicious instructions within the plan tasks.
  • Capability inventory: The subagents are given broad capabilities including file implementation, test verification, and git operations (commit), allowing for significant environment modification.
  • Sanitization: The skill lacks logic to validate or sanitize task content before it is passed to subagents for execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 08:51 AM