CLAUDE LABJP
CLI — Claude Code v2.1.261 landed on September 4. Of its 67 changes, 46 are fixes, and on the CLI side they cluster around input handling and background executionSKILLS — The new /skill-doctor lists which loaded skills go unused and how much context each one costs you every turn. As a starting point for cleanup, it is plainly usefulCOST — In one real-world setup the skill list alone consumed roughly 13,800 tokens per turn. Eighty-four skills had never been called, and duplicates from double-enabled plugins accounted for 41% of the totalLIMIT — New bashOutputMaxChars and taskOutputMaxChars settings raise how much command and background output reaches Claude inline before it is spilled to a file, up to 128K charactersBREAKING — Prompt word-editing keys now follow Bash and keybindingFlavor no longer has any effect. In auto mode, links that embed content into public diagram renderers count as uploadsSAFEGUARDS — Enterprise Frontier Safeguards was announced on September 1. The monitoring data sits in cloud infrastructure the customer controls rather than Anthropic's, and there is no extra chargeCLI — Claude Code v2.1.261 landed on September 4. Of its 67 changes, 46 are fixes, and on the CLI side they cluster around input handling and background executionSKILLS — The new /skill-doctor lists which loaded skills go unused and how much context each one costs you every turn. As a starting point for cleanup, it is plainly usefulCOST — In one real-world setup the skill list alone consumed roughly 13,800 tokens per turn. Eighty-four skills had never been called, and duplicates from double-enabled plugins accounted for 41% of the totalLIMIT — New bashOutputMaxChars and taskOutputMaxChars settings raise how much command and background output reaches Claude inline before it is spilled to a file, up to 128K charactersBREAKING — Prompt word-editing keys now follow Bash and keybindingFlavor no longer has any effect. In auto mode, links that embed content into public diagram renderers count as uploadsSAFEGUARDS — Enterprise Frontier Safeguards was announced on September 1. The monitoring data sits in cloud infrastructure the customer controls rather than Anthropic's, and there is no extra charge
TAG

context budget

1 articles
Back to all tags
Related:
Claude Code1Agent Skills1plugins1indie development1
Claude Code/2026-09-06Beginner

I counted 42 skills on my shelf, and four of them were just a note saying they had moved

Claude Code v2.1.261 added /skill-doctor. Before running it I counted my own skill shelf from the files, found four tombstones, and learned that the obvious awk one-liner for spotting duplicate plugin names quietly reports skills that are not duplicates at all.