irac-practice

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFE
Full Analysis
  • Local File Access: The skill is designed to interact with configuration and session files located in the ~/.claude/plugins/config/claude-for-legal/ directory. This behavior is used to customize the learning experience based on a student's classes and to track performance patterns over multiple sessions. These operations are localized and align with the skill's stated purpose.
  • Indirect Prompt Injection Surface: This skill processes user-provided essays and answers as its primary input. While this creates a surface for potential indirect prompt injection, the skill does not have access to sensitive capabilities—such as network operations or arbitrary command execution—that would make such an injection risky. This is a standard consideration for text-processing skills.
  • Legal Advice Safety Protocols: The instructions contain a 'Real-matter check' designed to identify when a user is seeking advice for actual legal issues rather than academic practice. It includes specific triggers and a mandatory disclaimer to ensure the agent does not perform unauthorized legal tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 09:46 PM
Security Audit — agent-trust-hub — irac-practice