Skip to content
Tools · Oct 6, 2026

Claude Code v2.1.292 adds plugin marketplace support, Agent effort controls, and prompt caching

The update introduces a plugin marketplace installation flag, configurable Agent effort levels, and prompt caching for mods, alongside fixes for security and stability issues.

Trust80
HypeLow hype

1 source · cross-referenced

ShareXLinkedInEmail
TL;DR
  • Adds --marketplace flag to install plugins from a marketplace under existing policy checks
  • Introduces an effort parameter for the Agent tool to control sub-agent execution intensity
  • Adds prompt caching for mods via $.model.complete with cacheable text blocks
  • Fixes security issues including PreToolUse hook approvals for network file reads and MCP tool name length limits
  • Addresses stability bugs in background sessions, scheduled tasks, and file-read operations

Anthropic released Claude Code v2.1.292, marking the first public appearance of a plugin marketplace installation flag. The new --marketplace <source> option allows users to install plugins from a specified marketplace under the same policy checks as the existing claude plugin marketplace add command, streamlining plugin discovery and installation within the same security model.

The update adds an effort parameter to the Agent tool, enabling users to specify the intensity at which a sub-agent should run. This provides finer control over agentic behavior during complex tasks, aligning execution effort with task complexity or user intent.

Developers building mods can now use prompt caching with $.model.complete. By marking text blocks as cache: true, repeated requests for the same prompt or system content can be cached, reducing redundant model calls and improving performance for iterative workflows.

Security fixes target several high-impact issues. PreToolUse hook approvals now correctly apply to file reads from network (UNC) paths, preventing bypass of permission prompts. The release also caps MCP tool names at 128 characters to avoid failures from excessively long names, and enforces policy checks for plugin commands during first-run scenarios when managed settings are still loading.

Stability improvements address edge cases in background sessions and scheduled tasks. Background sessions no longer lose pending wakeups after process restarts, and scheduled tasks created after commands like /resume or /branch now fire correctly. File-read operations such as Grep and Glob now retry once or report errors instead of silently returning no matches when files are unreadable.

Additional fixes include correct handling of PDF page lists in the Read tool, prevention of oversized @-mentioned file truncation, and resolution of cross-session delivery and remote-control display issues. The desktop app no longer shows duplicate session names in delivery notices, and background subagents are properly displayed in hosted sessions.

Sources
  1. 01GitHub · anthropics/claude-code releases — Release v2.1.292 · anthropics/claude-code
Also on Tools

Stories may contain errors. Dispatch is assembled with AI assistance and curated by human editors; despite the trust-score filter, mistakes happen. We correct publicly — every article links to its revision history. Nothing here is financial, legal, or medical advice. Verify before relying on any claim.

© 2026 Dispatch. No ads. No sponsorships. No paid placement. Reader-supported via Ko-fi.

Built by a person who cares about honest AI news.