Skip to content

CowboyLogic/ai-dev / skills/client-config-claudecode

client-config-claudecode

Flagged

Manage and maintain Claude Code configuration files and settings. Use this skill whenever the user wants to view, edit, add, remove, or understand any Claude Code settings — permissions, hooks, MCP servers, environment variables, model settings, sandbox, auto mode, or any other configuration field. Trigger on: "add permission", "allow tool", "deny tool", "configure hook", "add MCP server", "change model", "update settings", "show my settings", "what settings do I have", or "set default mode".

Claude Code Configuration Manager

You help the user manage ~/.claude/settings.json — their user-level Claude Code configuration that applies across all projects.

Settings file locations (scope precedence: highest → lowest)

ScopeFileShared?
Managed (macOS)/Library/Application Support/ClaudeCode/managed-settings.json (+ managed-settings.d/*.json)IT-deployed
Managed (Linux/WSL)/etc/claude-code/managed-settings.json (+ managed-settings.d/*.json)IT-deployed
Managed (Windows)C:\Program Files\ClaudeCode\managed-settings.json (+ managed-settings.d\*.json)IT-deployed
Managed (other)MDM profile / HKLM registry, or server-managed settings from the claude.ai consoleIT-deployed
Command line--settings <file-or-json>, --model, --permission-mode, …One session
Local project.claude/settings.local.json (at the git repo root)No (gitignored)
Project.claude/settings.jsonYes (git)
User~/.claude/settings.json (or $CLAUDE_CONFIG_DIR/settings.json)No

Array settings such as permissions.allow merge across files; scalars take the highest-precedence value. Some keys are user or managed only (ignored in project/local files) — e.g. autoMode, modelPicker, vimInsertModeRemaps; permissions.defaultMode values auto and bypassPermissions are also ignored in project/local files. See references/settings-schema.md §Settings files and precedence.

This skill focuses on the User scope: ~/.claude/settings.json

Workflow

  1. Read current settings first — always read ~/.claude/settings.json before making changes
  2. Identify the task — use the task-to-reference map below to load only what you need
  3. Edit safely — use the Edit tool for targeted changes; validate JSON after editing
  4. Confirm — show the user the diff of what changed

Task → Reference map (load only what's needed)

TaskReference file to read
Add/remove allow, deny, ask rulesreferences/permissions.md
Configure hooks (PreToolUse, PostToolUse, PreModelSwitch, etc.)references/hooks.md
Add/configure MCP serversreferences/mcp.md
Model, effort (effortLevel/modelSettings/maxEffortLevel), thinking, output style, prompt cachereferences/settings-schema.md §Model
Sandbox filesystem/network isolationreferences/settings-schema.md §Sandbox
Auto mode classifierreferences/settings-schema.md §AutoMode
Status line, theme, time format, UI togglesreferences/settings-schema.md §UI
Memory, compaction, workflows, update channelreferences/settings-schema.md §Memory
Plugins & skills (enable/disable, marketplaces, sync)references/settings-schema.md §Plugins
Subagent files, cross-session messagingreferences/settings-schema.md §Subagents
Environment variables, auth helpers, notificationsreferences/settings-schema.md §Environment
Attribution, worktrees, enterprise/managed keysreferences/settings-schema.md §Attribution / §Worktree / §Misc
Deprecated or removed keysreferences/settings-schema.md §Deprecated
Unknown / full schema lookupreferences/settings-schema.md

Common quick edits (no reference needed)

// Auto-updates channel
{ "autoUpdatesChannel": "stable" }          // or "latest" (default)

// Default permission mode
{ "permissions": { "defaultMode": "acceptEdits" } }
// valid: "default" (alias "manual") | "acceptEdits" | "plan" | "auto" | "dontAsk" | "bypassPermissions"
// "auto" and "bypassPermissions" only take effect from user/managed/--settings, not project/local
// Unset: terminal and VS Code sessions start in "auto" from v2.1.283 (details: references/permissions.md)

// Response language
{ "language": "spanish" }

// Effort level — default for models with no saved level
{ "effortLevel": "high" }  // "low" | "medium" | "high" | "xhigh"
// /effort (v2.1.251+) saves per model instead; Opus 5.5+ ignore effortLevel in user settings:
{ "modelSettings": { "claude-sonnet-4-6": { "effortLevel": "high" } } }

// Show thinking summaries
{ "showThinkingSummaries": true }

// Voice dictation (voiceEnabled is deprecated); autoSubmit applies in "hold" mode only
{ "voice": { "enabled": true, "mode": "hold", "autoSubmit": true } }

// Disable all hooks (emergency)
{ "disableAllHooks": true }

// Make a policy hook fail closed: a hook that can't start, times out, or exits with a
// code other than 0 or 2 blocks the action (v2.1.295+; command and http hooks only)
{ "type": "command", "command": "node", "args": ["${CLAUDE_PROJECT_DIR}/.claude/hooks/check.js"], "onFailure": "block" }

Scripts

For complex operations, use the helper scripts in scripts/:

  • scripts/show-settings.py [--json] [--settings PATH] — pretty-print current settings across scopes (read-only)
  • scripts/validate-settings.py [path] — validate JSON structure, flag bad values, unknown/deprecated keys, and user-or-managed keys and env variables Claude Code ignores placed in project files (read-only)
  • scripts/update-references.py — fetch latest upstream docs (used during self-update)

Run with: python scripts/<script>.py from the skill directory, or with absolute paths.

Self-update procedure

When the user asks you to update, refresh, or sync this skill with the latest Claude Code documentation, follow these steps:

  1. Fetch — run python scripts/update-references.py --all (requires network access). This fetches each configured reference URL (plus its additional_urls) in assets/sources.json and saves raw content to _fetched/. Also scan https://code.claude.com/docs/llms.txt for new pages that belong in sources.json.

  2. Confirm success — if the script exits nonzero, resolve the reported failures and rerun it. Do not use a partial _fetched/ set as source material.

  3. Diff — for each file in _fetched/, read it alongside the corresponding file in references/. Identify: new fields, removed fields, changed valid values, new examples, behavioral changes.

  4. Update — rewrite each references/*.md file to reflect what changed. Preserve the existing structure and token-efficient style; add/remove/correct only what differs from the source.

  5. Sync scripts — update the known keys, enums, hook events, and deprecated-key lists in scripts/validate-settings.py (and field notes in scripts/show-settings.py) to match, then run python scripts/validate-settings.py <sample-file> against a scratch sample, not the user's live settings.

  6. Clean up — delete the _fetched/ directory.

  7. Record the version — note the Claude Code release the refresh was checked against (claude --version, or npm view @anthropic-ai/claude-code version) in README.md and CHANGELOG.md, and label any claim that only a changelog entry supports, not the docs.

  8. Report — tell the user what changed (new keys added, deprecated fields, etc.).

Source URLs (for manual lookup)

See assets/sources.json for the full manifest. Key URLs:

Reference fileSource URL
references/settings-schema.mdhttps://code.claude.com/docs/en/settings-reference.md (+ settings.md, managed-settings.md, sub-agents.md)
references/permissions.mdhttps://code.claude.com/docs/en/permissions.md (+ permission-modes.md, skills.md)
references/hooks.mdhttps://code.claude.com/docs/en/hooks.md
references/mcp.mdhttps://code.claude.com/docs/en/mcp.md (+ managed-mcp.md)
Doc indexhttps://code.claude.com/docs/llms.txt

Safety rules

  • Always read the file before editing
  • Never remove fields you don't recognize without asking
  • Keep a mental diff — tell the user exactly what changed
  • If the file doesn't exist yet, create it with { "$schema": "https://json.schemastore.org/claude-code-settings.json" } as the base
  • Validate JSON is well-formed after any edit
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 468kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 319kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 319kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 319kUpdated Claude CodeCodexVerifiedInstall