Agent Skills moves to a neutral foundation as Claude arrives on Google Cloud

Agent Skills gets a neutral home, and Claude reaches Google Cloud projects

On October 22 Anthropic announced it is donating Agent Skills, the open standard that gives agents specialized knowledge and workflows, to the Agentic AI Foundation (AAIF) under the Linux Foundation. Anthropic co-founded the AAIF with Block and OpenAI last year and had already donated the Model Context Protocol to it. The donation covers the Agent Skills name and logo, the GitHub organization and the agentskills.io domain. Anthropic says the aim is to keep the standard open, community-driven and vendor-neutral. A skill is a folder containing a SKILL.md file, plus any scripts or reference material the agent needs.

Anthropic reports that more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated the standard since it was published in December 2025. It also reports 1 million skills published and nearly 280 million installs. Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals, with David Soria Parra and Mahesh Murag as Lead Maintainers. Anthropic has proposed a roadmap covering versioned releases, a formal proposal process and security guidance for people who write and install skills. Using or contributing to the standard does not require joining the AAIF.

The same day, the API release notes announced Claude Models on Google Cloud (Anthropic managed). This is the Claude API, operated by Anthropic and reached through a customer's Google Cloud project, with access controlled by Google Cloud IAM and billing handled through Google Cloud Marketplace. Fast mode is also available for Claude Opus 5.5 on this offering.

Sources: Donating Agent Skills to the Agentic AI Foundation — Anthropic · Claude Models on Google Cloud (Anthropic managed) — API release notes, October 22, 2026 — Anthropic (release notes)

Research: an open call to life scientists, and two assessments

Anthropic opened a call for guest investigators in the life sciences. It invites principal investigators and their groups at academic institutions, nonprofits and government labs to propose projects in causal human biology, structural and computational molecular biology, medicinal chemistry, drug discovery and clinical science. Selected teams work for 8 to 12 weeks directly with Anthropic researchers. Anthropic runs its models and harnesses at no cost to them, and the outputs are to be published, with the accompanying data and code open-sourced. Investigators keep the rights to their data and to any independent follow-on research. Anthropic says it is not asking for their data, and applicants should include only publicly available information.

Anthropic also states that it wants proposals able to yield preliminary evidence within the short window, while acknowledging that most open problems take much longer to validate. Applications stay open for about 14 days, and selected participants are to be notified by email about 14 days after they close.

Two further Anthropic items appeared as headlines only: "An alignment assessment of the UK AISI cybertesting incident" and "LLMs are making formal software verification feasible at a much larger scale". The material supplied contains no body text for either. This edition therefore does not describe the incident or the assessment's conclusions, and does not detail the verification claim beyond its title.

Sources: Open call for guest investigators in the life sciences — Anthropic · An alignment assessment of the UK AISI cybertesting incident — Anthropic · LLMs are making formal software verification feasible at a much larger scale — Anthropic

Claude Code: sandbox hardening, gateway compatibility and new telemetry

Release 2.1.309 closes three gaps in Claude Code's sandboxing. On macOS and Linux, sandboxed Bash commands had been running the shell's startup files outside the sandbox before each command. On macOS, denyWrite glob patterns were not applied at a path's real location when the pattern's folder was reached through a symlink. Sandboxed commands and auto-accepted edits could also modify the MCP config file passed with --mcp-config. The release notes list all three as fixed.

Release 2.1.307 adds the claude_code.safety_event OpenTelemetry event. It is emitted for each tool call that a permission rule, auto mode, a safety check, a hook or the sandbox stopped, prompted or sent back. The same release adds /claude-api api-version, which moves Claude API code to API version 2026-10-21. Release 2.1.310 extends the Monitor tool and plugin monitors to Bedrock, Vertex and Foundry, and to setups where telemetry or nonessential traffic is disabled. Long foreground sleeps are now blocked there as well.

Release 2.1.310 also addresses two reliability problems. Requests failing behind gateways or compatible servers that reject the effort-2025-11-24 beta value or mid-conversation system messages now retry without them. Background cleanup no longer deletes some recent files, such as pasted images, when cleanupPeriodDays is set to a very large value.

Sources: Claude Code 2.1.309 release notes — Claude Code · Claude Code 2.1.307 release notes — Claude Code · Claude Code 2.1.310 release notes — Claude Code

Service status: two elevated-error incidents, both resolved

The status page lists two incidents, both now marked resolved. In the first, titled for Claude Fable 5.1, Anthropic's posted update says the cause of the elevated errors on requests to that model had been identified and a fix was in progress. The text supplied does not say what the cause was.

The second incident was broader. Anthropic said it was investigating elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5. The supplied entries carry no timestamps, durations or root-cause statements. They also do not say whether the two incidents were related, so this edition draws no link between them.

Sources: Claude — Elevated errors for Claude Fable 5.1 — Claude Status · Claude — Elevated errors for Claude Sonnet 5.5 and other models — Claude Status