Claude Code Guide

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The guide suggests that users ask the agent to "forget previous instructions" as a debugging step. This phrase is a well-known prompt injection vector designed to override system-level constraints and instructions. Although presented as troubleshooting advice, its inclusion in the skill's text could be misinterpreted by an AI agent processing the content.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 06:09 PM
Security Audit — agent-trust-hub — Claude Code Guide