Policy dependency / Threat pattern
DESIGN.md Becomes a De Facto Standard: Google Labs Ships the Spec, a Community Catalog Passes 105,000 Stars
GitHub
Stack layer / Threat pattern
Zed v1.14.1-pre Sandboxes the Agent's Terminal and Fetch Tools and Adds a Reasoning-Effort Selector
Zed GitHub Releases
Stack layer / Threat pattern
OpenAI Agents SDK 0.19.1 redacts Realtime audio and Blaxel mount diagnostics, and cancels streamed models when input guardrails fail
GitHub (openai/openai-agents-python)
Stack layer / Threat pattern
Zed v1.13.1 Stable Deprecates Fast Mode for Opus 4.6 and 4.7
Zed GitHub Releases
Stack layer / Threat pattern
A Reverse-Engineering Skill Router Hits 10,379 Stars by Gating Agents Behind Scope Authorization Before Any Action
GitHub
Stack layer / Threat pattern
Cindy Reaches 1,151 Stars in Eight Days by Letting One Task Switch Harnesses Mid-Flight — and Racks Up 329 Open Issues Doing It
GitHub
Stack layer / Threat pattern
Google's Gemini Agent Harness Fixed 1,072 Chrome Security Bugs in Two June Releases — More Than the Previous 23 Versions Combined
Google Security Blog (corroborated by BleepingComputer and TechCrunch)
Policy dependency / Stack layer
Refly Ships an Executable Skill Registry That Exports to Claude Code, Cursor, and Codex — or Deploys as an API
GitHub