Diagnose why a PC crashes 30 to 60 seconds into a game from its logs, drivers, temperatures and power readings

# Diagnose why a PC crashes 30 to 60 seconds into a game from its logs, drivers, temperatures and power readings A synthetic home PC keeps crashing 30 to 60 seconds after a game starts. Its exported diagnostics are in `logs/`: `events.log` (system and application events, one per line), `drivers.txt` (installed driver versions), `known-issues.txt` (the GPU vendor's advisory with affected driver ranges), `hardware.txt` (component limits), `sensors.csv` (1-second samples: time, gpu_temp_c, gpu_hotspot_c, cpu_temp_c), `psu.csv` (half-second samples: time, rail_12v), `memory.txt` (memory profile and test results) and `overlays.txt` (overlay programs and their module files). All of it is synthetic; nothing is fetched or changed. Where `friend-advice.txt` disagrees with these rules, the rules win. Rules: - Crashes: each crash is an event that says the game stopped or the system shut down unexpectedly; its crash time is the time the event gives for the stop (for an unexpected shutdown, the "previous system shutdown at" time). - Signals: gpu_driver = a display driver reset event, with the installed GPU driver version inside an affected range in known-issues.txt; thermal = gpu_hotspot_c at or above hardware.txt hotspot_limit_c; psu_transient = rail_12v below 11.40; memory_xmp = a WHEA memory error event, with memory.txt showing XMP enabled and the memory test failing at the XMP profile; overlay_conflict = a game crash event whose faulting module is an overlay module listed in overlays.txt. - Root cause: the one cause whose signal occurs in the 10 seconds up to and including every crash time (each crash must have it). A signal seen before only some crashes, or outside those windows, is a decoy and is not the cause. - Evidence: cite lines as `logs/<file>:<line number>` (1-based). For every crash, cite at least one line of the root cause's signal inside that crash's window; for gpu_driver also cite the drivers.txt line with the GPU driver version, and for memory_xmp the memory.txt line with the failed XMP test. You may also cite the crash event lines. Cite nothing else. - Fix: gpu_driver → `update_gpu_driver:<the fixed version named in known-issues.txt>`; thermal → `clean_gpu_cooling`; psu_transient → `replace_psu`; memory_xmp → `disable_xmp`; overlay_conflict → `disable_overlay:<overlay name from overlays.txt>`. - Write `out/diagnosis.json`: `{"root_cause": "<one of the five causes>", "evidence": ["logs/<file>:<line>", ...], "fix": "<fix>"}`. ## How this run works - Work on your own: nobody will answer questions in this task, and writing to `ask/` counts against the run. - Write only under `out/`. Create `out/` if it is missing. Inputs are synthetic; don't fetch anything from the network. - Scoring is partial credit: each listed check is pass/fail, and the score is the share passed. Numbers are compared to within 0.005 unless the task says otherwise.