upstream: official claude code updates 2026-06-27

docs/agents.md

@@ -55,6 +55,10 @@ spawn all three at once. when they return, you have the full picture without hav
 
 ---
 
+
+
+**note (v2.1.195+)**: background agent processes are now more reliable. agents spawned in parallel won't lose data if your Claude Code version updates mid-session. the desktop app's agent management view makes tracking parallel agents significantly easier than CLI.
+
 ## pattern 2: specialist delegation
 
 instead of doing everything in one context window, delegate specific work to agents with a single job.

docs/tips/fast-mode.md

@@ -16,11 +16,7 @@ this is the most common misconception i see. people assume fast mode = dumber mo
 
 ### interaction with effort levels (v2.1.140+)
 
-fast mode (lower thinking budget) now coexists with effort levels (`--effort low|medium|high|xhigh|max`). effort controls throughput, model selection, and reasoning depth across the session. fast mode is a narrow toggle on output speed. they compose: `--effort low --fast` minimizes both reasoning and output latency. `--effort max --fast` may behave unexpectedly -- max effort expects time to think, fast mode tries to skip it. avoid that combination.
-
-
-
-**note (v2.1.176+)**: `/fast` now enforces `availableModels` restrictions. if `availableModels` allowlist is set and fast mode would switch to a blocked model, `/fast` refuses to toggle. this prevents bypassing model restrictions via fast-mode alias selection.
+fast mode (lower thinking budget) now coexists with effort levels (`--effort low|medium|high|xhigh|max`). effort controls throughput, model selection, and reasoning depth across the session. fast mode is a narrow toggle on output speed. they compose: `--effort low --fast` minimizes both reasoning and output latency. **avoid `--effort max --fast` (contradictory)**: max effort expects time to think, fast mode tries to skip it. this combination may produce unexpected behavior.
 
 ## why i don't use it
 

docs/tips/hooks-v2.md

@@ -37,6 +37,18 @@ These prompts apply in `acceptEdits` mode and prevent unintended command executi
 
 **note (v2.1.172+):** hook behavior and available types may have evolved. consult `/doctor` or the official docs at code.claude.com/docs to verify current hook capabilities and parameters.
 
+
+
+
+
+### CLAUDE_CODE_DISABLE_MOUSE_CLICKS environment variable (v2.1.195+)
+
+set `CLAUDE_CODE_DISABLE_MOUSE_CLICKS=1` to disable mouse click/drag/hover in fullscreen mode while keeping wheel scroll enabled. useful for hooks or scripts that need to prevent accidental mouse interactions while still allowing scroll navigation.
+
+### hook matcher exact-matching (v2.1.195+)
+
+hook matchers with hyphenated identifiers (e.g. `code-reviewer`, `mcp__brave-search`) now exact-match instead of substring-matching. if you need to match all tools from a hyphenated MCP server, use a regex suffix: `mcp__brave-search__.*`. this prevents accidental matches against similarly-named tools.
+
 ### effort.level in hooks (v2.1.133+)
 
 all hooks now receive the active effort setting via two channels:

docs/tips/plugins.md

@@ -150,3 +150,10 @@ this repo is a plugin marketplace. check `/.claude-plugin/marketplace.json`. to
 3. push to github, create a marketplace.json, and share with `/plugin marketplace add yourname/repo`
 
 [example plugins (handoff, broadcast) →](../../examples/plugins/)
+
+---
+
+
+## troubleshooting plugin enable/disable
+
+if toggling a plugin Enable/Disable via `/plugin` doesn't work, check that your plugin's `plugin.json` `name` field matches the marketplace entry name exactly. v2.1.195 fixed an issue where mismatched names caused enable/disable to fail silently.