Claude Code v2.1.198: Chrome GA, Agent Notifications, /dataviz, and AWS Gateway

Anthropic released Claude Code v2.1.198 with several quality-of-life improvements, new skills, and a handful of stability fixes. Here's what's relevant for daily use.
Chrome Extension Now GA
The Claude in Chrome extension is now generally available — no longer in preview.
Background Agent Notifications
Claude agents sessions that need input or finish now fire the Notification hook with agent_needs_input and agent_completed event types. This lets downstream scripts react to agent lifecycle changes.
New /dataviz Skill
The /dataviz skill provides chart and dashboard design guidance and includes a runnable color-palette validator — useful for teams generating visualizations via Claude.
Gateway: AWS Upstream Provider
Claude Platform on AWS (anthropicAws) is now available as an upstream provider in the Gateway. When a model is not found, the response advances the failover chain automatically.
Subagent and Workflow Improvements
- Background agents launched from
claude agentsnow commit, push, and open a draft PR when they finish code work in a worktree — instead of stopping to ask for confirmation. - The built-in Explore agent now inherits the main session's model (capped at Opus) instead of always running on Haiku.
- Subagents and context compaction now inherit the session's extended thinking configuration, improving output quality on delegated tasks.
- Subagents now treat messages from the agent that launched them as normal task direction.
- Agent teams: a teammate that dies on an API error now reports
failedto the lead, and messaging a stuck teammate wakes it to retry immediately.
Notable Fixes
- Network drop resilience: transient errors like
ECONNRESETnow retry with backoff instead of aborting the turn. - Background tasks in web, desktop, and VS Code task panels no longer get stuck on "Running" after completion or session resume.
- AWS STS token refresh: sessions using Claude Platform on AWS or Mantle no longer dead-end with
Please run /login—awsAuthRefreshruns automatically. - macOS background agents: fixed "no route to host" for local-network hosts by declaring Local Network entitlements.
- Highlight.js upgraded to v11 for improved syntax highlighting in code blocks, diffs, and file previews.
/branchnow derives its default fork name from the first real prompt instead of the compaction summary.
Removed: /agents Wizard
The interactive /agents wizard has been removed. Manage subagents by asking Claude directly or editing .claude/agents/ config files.
📖 Read the full source: GitHub Claude-Code
👀 See Also

GPU Power Consumption Deviates from Token Predictor Theory in Small LLMs
An experiment testing the 'stochastic parrot' theory on four 8B-parameter models found GPU power consumption often scales non-linearly with token count, with divergence rates ranging from 7.7% to 36.7%. The study also revealed persistent residual heat after philosophical queries and order-dependent effects.

OpenClaw users report high API costs from vague prompts, developer advises structured workflows
A Reddit user reports a $300 Anthropic bill from OpenClaw due to vague prompting, with the community noting the orchestrator works best with clear intentions and structured workflows rather than acting as a 'genie' for wishful thinking.

Domo CDO: Stop AI FOMO, Start with Spreadsheets
Domo chief design officer Chris Willis argues AI is being sold without a spec, creating fear-driven 'tokenmaxxing' theater. His fix: start by automating a spreadsheet process, not chasing moonshots.

OpenClaw Review: Reliability Issues in Current State, Value as Learning Tool
A developer with extensive AI platform experience reports OpenClaw struggles with reliability on basic multi-step tasks, making autonomous business applications questionable, but finds value in learning agent structure and orchestration.