Daily AI Builder Brief - May 27, 2026

Executive Summary

Claude Code 2.1.152 lands today with an actionable /code-review --fix and a new MessageDisplay hook for runtime UI control.
Skills can now declare disallowed-tools in frontmatter, giving authors a clean way to scope down agent tool access per skill.
/reload-skills and SessionStart reloading remove the restart penalty when iterating on skill files.
Auto mode no longer requires opt-in consent, and --fallback-model kicks in automatically when the primary model is missing.
Anthropic shipped a Compliance API plus 28 enterprise security integrations (CrowdStrike, Palo Alto, Okta, Zscaler, Wiz, Cloudflare) on May 26.
AWS MCP Server is now generally available, giving coding agents audited access to AWS services through the protocol.

Claude Code

Version 2.1.152 (May 27)

/code-review --fix now writes findings back into your working tree instead of just listing them, and /simplify is wired through the same path. The new MessageDisplay hook event lets you transform or hide assistant message text before render, which opens the door to redaction, tagging, and policy filters at the display layer. SessionStart hooks can also set sessionTitle and return reloadSkills: true, so session bootstraps can compose cleanly. Vim mode picked up reverse history search on / in NORMAL, and /usage now streams large session files. Source

Anthropic

Claude Compliance API and 28 Security Integrations (May 26)

The Compliance API exposes two streams to security teams: Claude Enterprise conversation content (chats, files, projects) and platform activity logs (logins, admin actions, config changes). It plugs into DLP, SASE, SIEM, identity, e-discovery, and AI observability stacks. If you've been waiting for procurement to clear Claude in a regulated environment, this is the unlock. Source

Korea Office (May 26)

KiYoung Choi appointed Representative Director ahead of the Seoul office opening. APAC expansion signal, not a builder-facing change today. Source

MCP Ecosystem

AWS MCP Server GA (May 2026)

AWS moved its MCP Server to general availability. Coding agents get secure, auditable access to AWS services without bespoke wrappers, and IAM stays the source of truth. Pair with the 2026-07-28 RC's stateless core and you can run remote MCP behind plain round-robin load balancers. Source

Broader AI

Nothing today.

What This Means For Builders

If you ship skills, audit frontmatter today: add disallowed-tools where a skill shouldn't touch shells or writes. It's a free security win.
The MessageDisplay hook plus /code-review --fix shifts review from advisory to applied. Wire it into your pre-commit flow.
Anthropic's Compliance API removes a real blocker for enterprise pilots. If you're selling into regulated buyers, refresh your security questionnaire answers this week.
AWS MCP GA means you can stop hand-rolling AWS tool servers. Migrate and inherit the audit trail.