feat(claude-code-settings): sync to Claude Code v2.1.76

[miteshashar]

Schema

• Add feedbackSurveyRate (number 0–1) for enterprise session quality survey control, as per settings documentation (v2.1.76)
• Add modelOverrides (object, string→string) to map Anthropic model IDs to provider-specific values like Bedrock ARNs, Vertex names, or Foundry deployments, as per model-config documentation
• Add worktree.sparsePaths (array of strings) for git sparse-checkout in large monorepos, as per settings documentation (v2.1.76)
• Add PostCompact hook event — fires after context compaction completes, as per hooks documentation (v2.1.76)
• Add Elicitation hook event — fires when an MCP server requests user input during a tool call, as per hooks documentation (v2.1.76)
• Add ElicitationResult hook event — fires after user responds to an MCP elicitation, as per hooks documentation (v2.1.76)
• Remove undocumented effortLevel default value ("high" was never documented as default); update description with tier-dependent defaults (Opus 4.6 defaults to medium for Max/Team) and /effort auto reference (v2.1.68)
• Update outputStyle description to mention custom styles support in ~/.claude/output-styles/ or .claude/output-styles/, as per output-styles documentation
• Update allowManagedDomainsOnly description to clarify that non-allowed domains are automatically blocked without user prompts (v2.1.74)
• Remove LS, MultiEdit, NotebookRead, Task from permissionRule pattern — unverified tools from original community PR Add Claude Code settings.json schema #4798, not found in tools-reference docs or ToolSearch. Also remove MultiEdit and Task(Explore) from permissionRule examples
• Add claudeMdExcludes — glob patterns for excluding CLAUDE.md files in monorepos, as per memory documentation. Not in changelog; documented feature missed in prior syncs.
• Add allowedHttpHookUrls — URL pattern allowlist for HTTP hooks, as per settings documentation. HTTP hooks added in v2.1.63; this security setting missed since then.
• Add httpHookAllowedEnvVars — env var name allowlist for HTTP hook header interpolation, as per settings documentation. Same origin as allowedHttpHookUrls.

Tests

• Add feedbackSurveyRate, modelOverrides, worktree to modern-complete-config.json for full coverage
• Add PostCompact, Elicitation, ElicitationResult hook examples to hooks-complete.json
• Add allowedHttpHookUrls and httpHookAllowedEnvVars to hooks-complete.json (from feat(claude-code-settings): sync to Claude Code v2.1.71 #5451 base)
• Add all 3 HTTP hook properties to modern-complete-config.json (from feat(claude-code-settings): sync to Claude Code v2.1.71 #5451 base)
• Changes made to pass coverage gate: 3 uncovered top-level properties (feedbackSurveyRate, modelOverrides, worktree)

Negative tests

• No changes needed — no new enums, patterns, or required fields added

Skipped

• KillShell in permissionRule pattern: not in tools-reference docs but verified present in tool harness — kept as undocumented internal tool
• $schema URL 301 redirect (http→https): SchemaStore CLI validator only accepts http://json-schema.org/draft-07/schema#
• Network env vars (HTTPS_PROXY, NODE_EXTRA_CA_CERTS, etc.): standard system/Node.js vars, not Claude Code settings.json properties
• No changelog-driven schema changes in v2.1.70–v2.1.71 (bug fixes only, from feat(claude-code-settings): sync to Claude Code v2.1.71 #5451)
• voice:pushToTalk (v2.1.71): keybinding, not a settings.json property (from feat(claude-code-settings): sync to Claude Code v2.1.71 #5451)
• Supersedes feat(claude-code-settings): sync to Claude Code v2.1.71 #5451 (v2.1.71 sync, now draft/closed)

🤖 Generated with Claude Code

[github-actions]

Thanks for the PR!

This section of the codebase is owned by @domdomegg, @bogini, and @ant-kurt - if they write a comment saying "LGTM" then it will be merged.