The Agent Skills Directory

[PROMPT_INJECTION]: The skill was analyzed for indirect prompt injection risks, as it processes configuration data and user-provided simulated scenarios. The skill effectively mitigates these risks by providing clear instructions to the AI and the user that all generated content is a non-authoritative starting point requiring verification and human review.
[DATA_EXFILTRATION]: The skill accesses a local configuration file at ~/.claude/plugins/config/claude-for-legal/legal-clinic/CLAUDE.md to load persona and jurisdictional context. This access is limited to the application's specific configuration directory and no network exfiltration patterns or unauthorized sensitive data access were observed.
[SAFE]: No obfuscation, remote code execution, persistence mechanisms, or privilege escalation attempts were found in the skill instructions.

ramp