Daily AI Builder Brief - June 5, 2026

Executive Summary

Claude Code shipped six releases this week, 2.1.160 through 2.1.165 (June 2 to 5).
New safety gate: Claude now prompts before writing to shell startup files like .zshenv and .bash_login.
Version control arrived: requiredMinimumVersion and requiredMaximumVersion managed settings let admins pin Claude Code versions.
Stop and SubagentStop hooks can now return hookSpecificOutput.additionalContext, so you can feed context back at turn end.
stdio MCP servers now receive CLAUDE_CODE_SESSION_ID on --resume, fixing session continuity for local servers.
Anthropic and MITRE published a year-long mapping of AI-enabled cyber threats.
Anthropic launched the Services Track and Partner Hub of the Claude Partner Network.
Reported across multiple outlets: programmatic and Agent SDK usage on subscription plans moves to a separate monthly credit pool on June 15.

Claude Code

Shell config write prompts (2.1.160)

Claude now asks before writing to shell startup files, and acceptEdits mode prompts before touching build-tool config. Fewer silent surprises in your dotfiles. Changelog

Version pinning for teams (2.1.163)

requiredMinimumVersion and requiredMaximumVersion managed settings let you enforce a version range across a fleet. Changelog

Hooks return context at Stop (2.1.163)

Stop and SubagentStop can return additionalContext, letting hooks inject state before the next turn. Changelog

Anthropic

A year of AI-enabled cyber threats, mapped with MITRE

Anthropic documented twelve months of AI-connected threats against the MITRE ATT&CK framework. Direct signal for anyone building offensive or defensive security tooling. Read

Claude Partner Network expands

New Services Track and Partner Hub for service partners building on Claude. Read

MCP Ecosystem

stdio servers keep session identity on resume (2.1.163)

Local stdio MCP servers now get CLAUDE_CODE_SESSION_ID on --resume, so server-side session state survives a resumed run. Changelog

MCP timeout floor fixed (2.1.162)

timeout config values below 1000ms were being floored; that's fixed, so sub-second timeouts behave as set. Changelog

Broader AI

Subscription programmatic usage moves to a credit pool (June 15, reported)

Multiple outlets report that programmatic and Agent SDK calls on Claude subscription plans shift to a separate monthly credit pool starting June 15. Not yet on the official newsroom; treat as reported, not confirmed. Report

What This Means For Builders

Audit any automation that writes dotfiles or build configs; the new prompts will pause non-interactive runs unless you adjust permissions.
If you run Agent SDK jobs on a subscription, model the June 15 credit-pool change into your budget before it bites.
Use additionalContext from Stop hooks to pass state between turns instead of bolting on external memory.