windbg

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). The required workflow runs WinDbg against a user-selected dump, process, or malware sample, and the agent reads strings and memory from it (e.g., da, du, s -a, TTD queries), so attacker-authored text inside that artifact can reach the LLM, but the attacker cannot push content into a queue the workflow monitors.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (low risk: 0.30). The skill is a standard WinDbg reference, but its kernel-debugging quick start tells the agent to run bcdedit commands that change the boot configuration and require administrator rights, which is a system-state modification.