Codex Knowledge Base
Configuration-Based Sandbox Escape: The Attack Class Every Codex CLI User Should Understand
In April 2026, Cymulate Research Labs published findings on a vulnerability class they termed Configuration-Based Sandbox Escape (CBSE).
Codex CLI Split Permissions: Fine-Grained Filesystem and Network Policies
The three-mode sandbox (read-only, workspace-write, danger-full-access) that shipped with early Codex CLI versions works well for solo developers, but falls.
End-to-End Testing with Codex CLI and Playwright: Agent-Driven Test Generation Pipelines
End-to-end test suites are the perennial bottleneck in modern development workflows.
Ticket-Driven Development with Codex CLI: Automating the Jira-to-Pull-Request Pipeline
The Atlassian Rovo MCP Server reached general availability in February 2026, exposing over 60 tools spanning Jira, Confluence, Bitbucket Cloud, Compass.
Codex CLI HIPAA Compliance in 2026: The Regulated Workspace Exclusion and What It Means
If your organisation processes Protected Health Information (PHI) and you are evaluating Codex CLI, there is a critical distinction buried in OpenAI's.
Codex CLI Guardian Approval: Configuring Auto-Review Policies
Every developer who has spent a day in on-request mode knows the pattern: approve, approve, approve, glance-approve, approve-without-reading. That reflexive.
Why Coding Agents Fail at Navigation (and How AGENTS.md File Maps Fix It)
Your coding agent can refactor a function, write tests, and call APIs — but ask it to find the right file in a monorepo.
What You Can Build with Codex CLI: 10 Real-World Setups from Simple to Advanced
Developers evaluating Codex CLI often ask the same question: what can I actually do with it? The answer depends on how deep you want to go. Codex CLI scales.
Using Codex CLI to Improve Published Algorithms: A Two-Stage Pipeline
A recent paper by Suwannik (April 2026) demonstrates something that should give every research-oriented developer pause.
The Harness Effect: Why the Same Model Scores 16 Points Higher in a Different Tool
Claude Opus running inside Cursor scores 93% on Terminal-Bench 2.0. The same model running inside Claude Code scores 77%. That is a 16-point differential.