design-review

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (SAFE): The skill uses a persona ('Rams') and instructions tailored specifically for design reviews. No commands to bypass safety filters, ignore previous instructions, or extract system prompts were detected.
  • [DATA_EXFILTRATION] (SAFE): The skill does not contain hardcoded credentials, sensitive file paths, or network operations targeting external domains.
  • [REMOTE_CODE_EXECUTION] (SAFE): There are no patterns involving downloading scripts (curl/wget) or executing remote code. No external dependencies are requested.
  • [Indirect Prompt Injection] (LOW): The skill ingests untrusted source code for review and possesses the capability to modify files if the user accepts a fix. This creates a surface for indirect prompt injection.
  • Ingestion points: File content accessed via $ARGUMENTS or user-selected component files.
  • Boundary markers: Absent. The instructions do not specify delimiters to isolate external code from the agent's internal instructions.
  • Capability inventory: File-read (to perform analysis) and file-write (via the offer to 'fix the issues directly').
  • Sanitization: None. The agent processes the code snippets directly as provided in the file stream.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:45 PM