design-it-twice

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs sub-agents to process project source code and grounding files to inform architectural designs. This pattern allows untrusted project content to enter the agent's context, which could lead to indirect prompt injection if those files contain instructions targeting the LLM.
  • Ingestion points: 'Grounding' file paths provided in SKILL.md and instructions to 'Explore the code further' in runner.md.
  • Boundary markers: None identified to delimit untrusted data or warn agents about embedded instructions.
  • Capability inventory: The skill instructions limit sub-agents to generating design proposals and sketches rather than executing code or performing system operations, which reduces the potential impact of an injection.
  • Sanitization: No sanitization or filtering of external content is performed.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 12:51 AM