CSP Allow-list Experiment
Summary
A web tool experiment demonstrating how to handle Content Security Policy errors in sandboxed iframes by intercepting fetch requests and prompting users to whitelist domains. The tool was built using GPT-5.5 via the Codex desktop app.
View Cached Full Text
Cached at: 05/13/26, 08:10 AM
Similar Articles
External sandboxes for coding agents
Simon Willison and Prime Radiant introduce the 'external sandbox' pattern for safely executing code from coding agents, releasing three sandbox server implementations (v86, container, and Quicksand-based) with a shared localhost API and network allow-list proxy for tools like Pi, Claude Code, and Codex.
webframe: evaluated as a scraper and a sandbox
Simon Willison publishes a research report evaluating webframe, Jeff Lindsay's lightweight Go tool for running web pages like programs in a headless system web view, finding it fast and comparable to shot-scraper but with rough edges like CSP failures and missing HTTP metadata.
I built a Windows-native security sandbox for local LLM agents. It treats tool calls like syscalls and default-denies high-risk actions. (Open Source, AGPL-3.0)
The author released Cyber-Butler-Security (CBS), an open-source Windows-native OS-level sandbox for local LLM agents that treats tool calls like syscalls, uses dual-account ACL isolation and a TOTP broker for high-risk actions, and fails closed by default.
Building a UMatrix Replacement
This article explores options for replacing the deprecated uMatrix browser extension under Chrome's Manifest V3, proposing a solution using declarativeNetRequest and Content Security Policy to control site permissions and subresource requests.
Universal XSS in uBlock Origin from website-driven element picker
A researcher disclosed a one-click universal XSS vulnerability in uBlock Origin's element picker that let malicious websites inject trusted filters and run arbitrary JavaScript on any site; the issue is now fixed and was largely discovered with the help of GPT-5.6 Sol.