Windows script debugger questions, answered one by one

A windows script debugger is a debugger you use to step through script code on Windows, set breakpoints and inspect variables. StatusRivet does not replace one. It runs your script, your build command and your test command through Shell, reads the error text and stack traces they print, and drafts a diff for the file that failed. Stepping stays in your own debugger, so use both.

A PDB file maps addresses to function names, source files and line numbers, so a stack trace reads as code instead of numbers. Symbols matter. Build with debug symbols and keep the matching PDB for every build you ship, because a mismatched PDB gives wrong function names. Without symbols a call stack shows addresses only. StatusRivet can search saved stack text, but it cannot recover names the PDB lacks.

The exception code names the failure: 0xC0000005 is an access violation, which means reading or writing invalid memory, and 0xC00000FD is a stack overflow. The call stack of the faulting thread then shows which function led to it. A windows 11 bsod with code 1a follows the same idea at system level: a stop code in a dump, read from the top frame down. Let grep find it.

The JIT setting, short for just-in-time, makes Windows start a debugger automatically when an application crashes. Lab machines only. It is a system setting, so change it only on a test machine, write down the previous value and restore it afterwards. StatusRivet can run the command you give it to read the value and save it to a text file first, and Ask first pauses before any change.

Kernel-level work is for drivers you write. Lab only. It normally needs two machines or a virtual machine: one runs the debugger, the other is the test target, linked by a debug connection. Never use your daily PC as the target. An irql not less or equal bsod is a typical stop code there, and the stack names the faulting driver function.

Yes, StatusRivet runs the tools itself. Cards show everything. Every action appears in the chat as a card with its real result, and an expanded card shows the full output, either the command result or the file diff. A chat assistant describes what you could do; this agent reads the file, runs the command and reports the actual exit code. You still approve changes in Ask first mode.

The file tools read_file, write_file, append_file, edit_file, delete_file, make_dir, move_file, copy_file, list_dir, grep and glob work inside the chat's working folder. Shell runs terminal commands such as your build and test commands and judges success by the real exit code; dangerous commands are blocked. Scope is yours. Shell, web and SSH are not limited to that folder, so you keep the work within your own systems.

The approval modes are Auto-run, Ask first and Plan only. Auto-run acts at once. Ask first pauses before actions that change something, such as writing a file, and shows the diff for you to accept. Plan only writes a plan and runs nothing, no commands and no file changes; you switch mode to carry it out. Plans survive restarts, so you can come back to one later.

Yes, windowsdebugging.com is the official StatusRivet website, and the download button here gives the original setup program. Compare its SHA-256 checksum with the published manifest before you run it. The current version is 2.9.26, released on 2 Nov 2026, and the setup program is about 34 MB. If anything looks wrong, use the Send feedback button inside the app.

No. StatusRivet shows real output, but it cannot promise that a diagnosis or a fix is correct. It works from text output, and a log with missing symbols leads to weaker answers. You read every diff, rebuild, and rerun your tests before you keep a change. An error kernel power event 41 in the log, for example, tells you the PC restarted unexpectedly, not which code caused it.

The model runs where you point StatusRivet. A cloud AI provider runs on the provider's servers through a standard API endpoint, a local model runs on your PC, and the built-in model is ready to use without your own endpoint. Test before trusting. The Test button checks an endpoint and shows the response time and the context window, so you can compare choices before a long debugging session.

The answer depends on the model, so it is not a bare yes or no. With a cloud AI provider, the prompt, including the text of any file the agent reads for that step, goes to that provider. With a model on your PC, nothing leaves the machine. Keys stay local. Chats and API keys stay on the PC either way, and secrets in tool output are redacted.

The download button gives the setup program of about 34 MB, not the portable .exe. The installer is not code-signed yet, so run certutil -hashfile with the path of the downloaded file and SHA256, and compare the result with the published manifest. Verify before running. Windows SmartScreen may warn; choose More info and then Run anyway only after the checksum matches.

The button records the time, your IP address, the browser user agent and the referral source, if there is one, to count downloads, and then redirects you to the installer. The website itself is made of static pages; hosting and a CDN process IP addresses and access logs, and a visit counter from statinside.com records page visits. The app has no telemetry.