PS

Prompt Crime Scene Investigator (PCSI) is a tool that helps teams understand why an AI agent failed. As AI systems become more autonomous, problems like prompt injection, wrong tool usage, memory issues, hallucinations, and multi-agent failures become difficult to debug. Instead of just showing logs, PCSI investigates the failure like a digital detective. Users upload logs, prompt traces, error outputs, or execution files, and the system automatically runs multiple specialist agents to analyze what went wrong. One agent collects evidence and creates a timeline, another checks for prompt injection, another analyzes reasoning mistakes, and another investigates tool misuse and coordination failures. Finally, the system creates a clear Crime Scene Report showing the root cause, timeline of events, confidence score, failure heatmap, and recommended fixes. Our goal was to make AI failures easier to understand and fix through an autonomous workflow that goes beyond a single prompt.