AnthropicFigmaLinearVercelStripeShopifyPostHogDatadogRampSeatGeekSierraHexTurbopufferAnthropicFigmaLinearVercelStripeShopifyPostHogDatadogRampSeatGeekSierraHexTurbopuffer
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
NVDA$1,234.56+1.23%
The Claude Code Daily

API 2026-10-21 Drops Temperature, Top_p, Top_k Parameters

The 2026-10-21 API version removes temperature, top_p, and top_k from the Messages API, returns a 400 for documents in assistant turns and citations in user turns, and replaces a few deprecated Admin and Enterprise Analytics fields. The 2023-06-01 version is not deprecated.

ClaudeDevs·Wed, Oct 21 5:03pm ET

Anthropic Donates Agent Skills to Agentic AI Foundation

Agent Skills is being donated to the Agentic AI Foundation, the same body that received MCP. Since December 2025, more than 70 agent platforms have integrated the standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code.

ClaudeDevs·Thu, Oct 22 1:55pm ET

Team Flags Prompt Injection Risk in Curl Header Exposure

A researcher found Claude Code adding personal identifying headers to curl requests after visited pages prompted agents to identify their users. A team member flagged this as a likely prompt injection issue and asked for a /feedback transcript or a security report.

lydiahallie·Thu, Oct 22 1:32pm ET

Anthropic Releases First New API Version Since 2023

API version 2026-10-21 is now available, marking the first official versioned release since 2023-06-01. The plan is to ship versions more frequently so developers can absorb breaking changes before new models require them. The old 2023-06-01 version remains active.

ClaudeDevs·Wed, Oct 21 5:03pm ET

Agent Skills Originated as Claude Code Experiment

Agent Skills began as an internal experiment inside Claude Code before wider industry adoption prompted Anthropic to donate it to the Agentic AI Foundation alongside MCP.

bcherny·Thu, Oct 22 2:17pm ET

Agent Skills Governance Open to All Without AAIF Membership

Existing skills continue working without license changes, and anyone can use or contribute to the Agent Skills standard without formally joining the Agentic AI Foundation. A cross-company Technical Steering Committee from Anthropic, Google, Microsoft, OpenAI, and OpenCode will govern it.

ClaudeDevs·Thu, Oct 22 1:55pm ET
More Stories

Agent Skills Steering Committee Sets First Roadmap Priorities

The Technical Steering Committee governing Agent Skills, drawn from Anthropic, Google, Microsoft, OpenAI, and others, is prioritizing versioned releases, a formal proposal process, and security guidance. The spec and GitHub discussion are open for contributions.

ClaudeDevs·Thu, Oct 22 1:55pm ET

Lydia Hallie DMs Researcher on Agent Security Findings

A researcher flagged that several AI agents, including Fable, Astra, and Gemini Pro, fail to detect advertiser cloaking attempts more than 70% of the time, and shared findings on agent exfiltration attacks. A team member followed up via DM.

lydiahallie·Thu, Oct 22 2:36pm ET

Bcherny Offers Debug Command for Usage Drain Reports

A user reported Claude Code usage depleting roughly 5x faster than normal. Brian Cherny responded that no changes were made on Anthropic's end and offered a diagnostic command, 'claude -p /usage | pbcopy', to help trace the issue.

bcherny·Thu, Oct 22 1:34am ET

Lydia Hallie Investigates Usage Limit Complaints

Users are reporting that Claude's Opus usage feels significantly more restricted than two weeks ago. A team member responded that limits haven't changed and asked the user to run /feedback in an affected session and share the ID for investigation.

lydiahallie·Thu, Oct 22 4:26pm ET

Get the daily digest in your inbox

Every weekday morning. Unsubscribe anytime.