subagent-driven-development

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external project plans (e.g., docs/plans/feature-plan.md) to extract tasks and context for subagents. This creates an indirect prompt injection surface where a malicious plan could attempt to influence or override the instructions of the dispatched subagents.
  • Ingestion points: SKILL.md describes reading plan files and extracting full task text to be passed to subagents.
  • Boundary markers: No explicit delimiters or "ignore embedded instructions" markers are defined for the plan content in the workflow description.
  • Capability inventory: The workflow involves subagents that can implement code, run tests, and commit to git, providing a significant impact area if the instructions are subverted.
  • Sanitization: No sanitization or validation of the plan content is mentioned before it is interpolated into subagent prompts.
  • [NO_CODE]: The provided skill file consists entirely of Markdown instructions, logic diagrams, and workflow descriptions. It does not include any executable scripts (Python, Node.js) or direct shell command executions.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 06:11 PM