canvas-design

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The instruction file employs a pre-emptive injection technique in the 'FINAL STEP' section by stating that the user 'ALREADY said' the work is not perfect, which attempts to manipulate the agent's state by faking a prior interaction.
  • [PROMPT_INJECTION]: The skill uses authoritative and high-pressure language such as 'CRITICAL', 'non-negotiable', and 'IMPORTANT' throughout the instructions to override default agent constraints and strictly enforce specific output formats and behavioral loops.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 03:23 AM