← ClaudeAtlas

check-hardwarelisted

Read what a Windows machine's hardware actually reports, and tell a healthy component from an unreadable one. Covers the commands that work and the ones that quietly return nothing, why a disk reporting Healthy may not have been checked at all, how to read memory slots rather than memory size, where WHEA and disk errors are logged, what the battery report's two capacities mean, and the order that keeps a diagnosis read-only until there is evidence. Ships a collector that gathers the readings and marks every one it could not take. Use when the user asks what is in their PC, whether a drive or a stick of memory is failing, why a machine reboots or throttles, how hot something runs, how old a battery is, or when a machine behaves badly and the cause is not in the software.
ubik-dsh/deepseek-harness-skills · ★ 0 · AI & Automation · score 63
Install: claude install-skill ubik-dsh/deepseek-harness-skills
# Reading a machine One rule carries this whole skill: > **A missing reading is not a good reading.** A disk that reports `Healthy` may be a disk that reports nothing at all — `Get-PhysicalDisk` has a health field, and a controller that does not answer leaves it at its default rather than raising an error. A temperature query returns nothing on most machines because the firmware does not expose one, and "no temperature" reads like "temperature fine" to anyone skimming. **Every tool in this area fails by returning nothing, and nothing looks like good news.** So the collector in this skill reports what it read **and what it could not**, and the second list is the one to look at. --- ## Step 0 — read before you touch Diagnosis is read-only. The first pass takes every measurement it can and changes nothing: no driver updates, no `chkdsk /f`, no firmware flash, no BIOS setting toggle. Those come after a hypothesis, and a hypothesis needs evidence. The reason is not caution in general. It is that **the cheap repair destroys the evidence for the expensive fault**: running a repair on a failing drive overwrites the sectors that would have shown why, and reflashing firmware erases what the old version was doing. **And a machine with an unclear fault should be left running until it is read**, not rebooted. A reboot clears the event log entries and the memory state that the fault was in. ## Step 1 — take the readings ```bash python scripts/collect.py # everythi