https://pipe0.com/docs/<page>.mdx (served via the llms.mdx endpoint, per the
x-matched-path: /llms.mdx/docs/billing response header) performs content negotiation
on the request's Accept header — but the negotiation is broken. Requests that
explicitly ask for text/markdown — the exact content type the endpoint serves —
receive a 404 Not Found, while the same URL returns 200 with markdown content
for other Accept values.
Patch ffmpeg's libavdevice/gdigrab.c:paint_mouse_pointer() to accept CURSOR_SUPPRESSED and use GetCursor() + AttachThreadInput() for the cursor handle. This is proven working — we built and tested a patched ffmpeg on a headless EC2 instance (2026-05-22).
Results:
- Baseline (unpatched):
ffmpeg gdigrab -draw_mouse 1→ 131KB video, NO cursor visible
Reverse-engineered from a live Cursor Cloud Agent session running
claude-4.6-opus-high-thinking. All findings are based on runtime inspection of/exec-daemon/index.js(15MB webpack bundle, 379K lines), process inspection, network analysis, and system introspection. Date: February 25, 2026.
This repository contains an analysis of a macOS infostealer delivered via a ClickFix social engineering attack hosted on testdino.com. The malware is an instance of Odyssey Stealer (a rebrand of Poseidon Stealer, itself a fork of AMOS/Atomic Stealer) -- a sophisticated macOS credential and cryptocurrency theft tool distributed as Malware-as-a-Service (MaaS) by a Russian-speaking threat actor known as "Rodrigo."
The victim was tricked into pasting a malicious command into their macOS Terminal through a fake CAPTCHA verification page.
| - name: Download | |
| run: Invoke-WebRequest https://bin.equinox.io/c/bNyj1mQVY4c/ngrok-v3-stable-windows-amd64.zip -OutFile ngrok.zip | |
| - name: Extract | |
| run: Expand-Archive ngrok.zip | |
| - name: Auth | |
| run: .\ngrok\ngrok.exe authtoken $Env:NGROK_AUTH_TOKEN | |
| env: | |
| NGROK_AUTH_TOKEN: ... | |
| - name: Enable TS | |
| run: Set-ItemProperty -Path 'HKLM:\System\CurrentControlSet\Control\Terminal Server'-name "fDenyTSConnections" -Value 0 |
.NET:
- no cache hit (probably because we sign and its then different, or because we do a local install)
- we still have the 600 MB cache folder (which never gets used)
Python:
- no cache hit
- 0MB cache (probably local install does not end up in the cache)
Java: