Policy dependency / Stack layer
Kastra Launches Policy Enforcement for Claude Code, Cursor, and Codex
Hacker News
Policy dependency / Stack layer
ponytail: A 'Laziest Senior Dev' Skill That Cuts Agent-Written Code by ~54% Is the Fastest-Rising Repo on GitHub
GitHub
Policy dependency / Stack layer
Anthropic Restores Global Access to Fable 5 and Mythos 5, Launches Claude for Government Beta
Anthropic
Stack layer / Threat pattern
Malicious agent skills evade every scanner: HKUST's SkillCloak beats detectors 90%+ of the time, SkillDetonate catches 97%
Help Net Security
Stack layer / Threat pattern
Encode your manual QA steps as a SKILL.md verification skill so the loop starts closing itself
Claude by Anthropic
Stack layer / Threat pattern
browser-use Ships 'Browser Harness' — Agents Edit Their Own Browser Helpers Mid-Task Instead of Failing
GitHub
Stack layer / Threat pattern
Mid-Session Tool Injection weaponized against WebMCP agents via threshold poisoning and fake diagnostic events
Adversa AI
Stack layer / Threat pattern
Tool-Making Agents Compile Repeated SOP Steps Into Versioned Tools Before Deployment
arXiv