upstream: official claude code updates 2026-05-20

[github-actions]

upstream changes detected

sources: anthropic/official-docs, shanraisshan/claude-code-best-practice
changes found: 11
files edited: 7
staleness flags: 10

what changed upstream

• [commit] docs(claude-settings): log v2.1.145 changelog entry (25 items) (link)
• [commit] docs(claude-settings): update to v2.1.145 (link)
• [commit] docs(claude-commands): log v2.1.145 changelog entry (link)
• [commit] docs(claude-commands): update to v2.1.145 (link)
• [commit] docs(development-workflows): log 2026-05-21 update (17 items) (link)
• [commit] chore(development-workflows): refresh stars + counts in workflows table (link)
• [commit] docs(concepts): add checklist rule upstream: claude code updates 2026-03-18 #8 (Location column factual accuracy) (link)
• [commit] docs(concepts): log 2026-05-21 run (v2.1.145) — Checkpointing location fix (link)
• [commit] fix(readme): Checkpointing location is file-edit tracking, not git-based; bump badge v2.1.145 (link)
• [commit] chore(changelog): log v2.1.145 skills run (3 new bundled skills) (link)
• [docs_change] Official Claude Code docs updated (link)

staleness flags

file	claim	reality	severity
docs/tips/fast-mode.md	fast mode does not switch to a cheaper or smaller model	v2.1.145 adds CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE env var, suggesting fast mode behavior may be configurable per deployment	low
docs/tips/monitor.md	monitor shipped april 9 2026, requires v2.1.98+	v2.1.145 is now current; monitor is stable and mature across all current versions	low
docs/tips/hooks-v2.md	five handler types listed (command, http, prompt, agent, mcp_tool)	v2.1.145 adds new hook event types (Stop, SubagentStop with background_tasks/session_crons context) and effort.level support in hook input	medium
docs/tips/settings-hierarchy.md	three levels: global, project, local	v2.1.145 adds claudeMdExcludes, autoMemoryEnabled, worktree.bgIsolation settings; settings count jumped from 60+ to 80+	medium
docs/tips/context-management.md	compaction cliff data from sessions, 32% of 30-60min sessions compact	autoMemoryEnabled (v2.1.145) may change compaction behavior by auto-preserving session memory; data may be stale	medium
docs/tips/subagents.md	worktree isolation uses default branching from origin/	v2.1.145 adds worktree.bgIsolation setting for background worktree isolation; worktree.baseRef changed default to 'fresh' in v2.1.133	medium
docs/cost.md	pricing snapshot from may 2026	v2.1.145 is now current (may 2026); official pricing should always reference docs.anthropic.com, not cached snapshots	low
docs/tips/mcp-integration.md	MCP servers are optional, most people never need any	v2.1.145 expands MCP integration; antigravity comparison shows MCP as native feature for google cloud; landscape may be shifting	low
docs/comparisons/antigravity.md	antigravity's post-preview pricing is unconfirmed as of may 2026	document correctly flags this as verify-at-purchase; no change needed but review antigravity.google/pricing before recommending	low
docs/tips/plan-mode.md	plan mode is a toggle with /plan command	v2.1.145 adds xhigh effort option; plan mode behavior may differ by effort level	low

proposed edits

• docs/tips/hooks-v2.md: v2.1.145 adds new hook fields for background task awareness; document this capability
• docs/tips/settings-hierarchy.md: v2.1.145 adds 20+ new settings; these three change compaction, memory, and worktree behavior meaningfully
• docs/tips/context-management.md: v2.1.145 adds auto-memory preservation which changes compaction dynamics; existing data may be from before this feature
• docs/tips/subagents.md: v2.1.145 adds background worktree isolation control; affects how parallel agents see parent session changes
• docs/tips/fast-mode.md: v2.1.145 adds a new env var that hints at future fast-mode customization; users should know this exists
• docs/cost.md: update last-verified date to may 21 2026; pricing table itself should never be cached, always link to official source
• docs/session-workflow.md: v2.1.145 auto-preserves memory; compaction behavior changes, existing workflow advice is stale
• docs/tips/hooks-v2.md: v2.1.145 adds xhigh effort level; hooks that handle effort need this new case

review checklist

• staleness flags accurate?
• version numbers and pricing correct?
• proposed edits read naturally?

auto-generated by official-watcher