subagent-driven-development

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its handling of untrusted implementation plans.
  • Ingestion points: Implementation plans such as docs/plans/feature-plan.md are ingested, and task text is extracted for subagent prompts in implementer-prompt.md and spec-reviewer-prompt.md.
  • Boundary markers: The skill uses Markdown headers to separate task text but lacks strong delimiters or instructions for subagents to ignore embedded commands within the untrusted content.
  • Capability inventory: Subagents have the ability to implement code, run tests via general-purpose tools, and perform git commits.
  • Sanitization: No sanitization or validation is performed on the plan text before it is interpolated into the prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 04:50 AM