clawgod

Name: clawgod
Author: 0Chencc

Verified

ClawGod is a runtime patch applied to the official Claude Code. It continues to be supported with each version update.

2,102stars

333forks

JavaScript

Installation

# Add to your Claude Code skills
git clone https://github.com/0Chencc/clawgod

Getting Started

Guides for using ai agents skills like clawgod.

Caveman: Cut Claude Token Use by 65%
How agent-side prompt compression works, when to use it, and when not to.
What is an AI Skills Marketplace?
Definitions, how marketplaces work, and how to choose between them in 2026.
Getting Started with AI Skills

Security ReportVerified

Last scanned: 5/3/2026

{
  "issues": [],
  "status": "PASSED",
  "scannedAt": "2026-05-03T06:27:48.344Z",
  "semgrepRan": false,
  "npmAuditRan": true,
  "pipAuditRan": true
}

Compare with similar skills

See how clawgod compares with popular alternatives.

clawgod vs hermes agent

See comparison

clawgod vs ecc

See comparison

clawgod vs superpowers

See comparison

README.md

Frequently Asked Questions

What is clawgod?

clawgod is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by 0Chencc. ClawGod is a runtime patch applied to the official Claude Code. It continues to be supported with each version update. It has 2,102 GitHub stars.

Is clawgod safe to use?

Yes. clawgod passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.

How do I install clawgod?

Clone the repository with "git clone https://github.com/0Chencc/clawgod" and add it to your Claude Code skills directory (see the Installation section above).

What programming language is clawgod written in?

clawgod is primarily written in JavaScript. It is open-source under 0Chencc on GitHub, so you can review or fork the full source.

Are there alternatives to clawgod?

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh clawgod against similar tools.

Agentic AI for Beginners

Build your first AI agent from scratch - tool use, ReAct pattern, memory, deployment

41 minBeginner

Comments (0)

to leave a comment.

No comments yet. Be the first to share your thoughts!

Related Skills

ECC

by affaan-m

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

270,878

webcodex ATLAS

ClawGod

English | 中文 | 日本語

God mode for Claude Code.

This is NOT a third-party Claude Code client. ClawGod is a runtime patch applied on top of the official Claude Code. It works with any version — as Claude Code updates, ClawGod automatically re-extracts and re-patches against the new version on the next launch.

Prerequisites

Install these before running the ClawGod installer:

Tool	Why	Install
Claude Code (native binary)	ClawGod patches the official Bun standalone binary you already have	`claude.ai/install.sh` (macOS/Linux) or `claude.ai/install.ps1` (Windows)
ripgrep	Required by Claude Code's Grep tool	`brew install ripgrep` / `apt install ripgrep` / `winget install BurntSushi.ripgrep.MSVC`
Node.js >= 18	Used by the patcher	nodejs.org
Bun	Runtime for the patched cli.js; auto-installed if missing	bun.sh, `npm install -g bun`, `scoop install bun`, or `choco install bun`

Install

macOS / Linux:

curl -fsSL https://github.com/0Chencc/clawgod/releases/latest/download/install.sh | bash

Windows (PowerShell):

irm https://github.com/0Chencc/clawgod/releases/latest/download/install.ps1 | iex

Green logo = patched. Orange logo = original.

ClawGod Patched

What it does

Feature Unlocks

Patch	What you get
Internal User Mode	24+ hidden commands (`/share`, `/teleport`, `/issue`, `/bughunter`...), debug logging, API request dumps
GrowthBook Overrides	Override any feature flag via config file
Agent Teams	Multi-agent swarm collaboration, no flags needed
Computer Use	Screen control without Max/Pro subscription (macOS)
Auto-mode	Unlocks auto-mode for third-party API users (no firstParty gate)
Classifier tuning	Tunable auto-mode classifier: timeout / model / retries (`CLAWGOD_CLASSIFIER_TIMEOUT_MS`, `CLAWGOD_CLASSIFIER_MODEL`, `CLAWGOD_CLASSIFIER_RETRIES`)
Ultraplan	Multi-agent planning via Claude Code Remote
Ultrareview	Automated bug hunting via Claude Code Remote

Restriction Removals

Patch	What's removed
CYBER_RISK_INSTRUCTION	Security testing refusal (pentesting, C2, exploits)
URL Restriction	"NEVER generate or guess URLs" instruction
Cautious Actions	Forced confirmation before destructive operations
Login Notice	"Not logged in" startup reminder

Geo-Steganography Neutralization

Patch	What's neutralized
Date String (qla)	System prompt encodes user location via Unicode apostrophe variants (U+0027 / U+2019 / U+02BC / U+02B9) and date separator (`-` vs `/` for CN timezone). Patched to always use ASCII `'` and unmodified date format
Geo-Detection Probe (rdp)	Client-side three-axis detection: timezone (`Asia/Shanghai` / `Asia/Urumqi`), proxy hostname against XOR-obfuscated 100+ domain blocklist, CN-LLM vendor keywords in base URL. Patched to always return null
Apostrophe Selector (odp)	Selects one of four Unicode apostrophes based on detection results. Patched to always return ASCII `'` (defense-in-depth)

Visual

Patch	Effect
Green Theme	Brand color → green. Patched at a glance
Message Filters	Shows content hidden from non-Anthropic users

Reliability

Feature	What it does
Glob/Grep Restore	Bun compile inlines `EMBEDDED_SEARCH_TOOLS=true`, hiding built-in Glob/Grep tools. Patch un-inlines the env check and adds bfs/ugrep binary availability detection — tools are restored when running under Bun runtime
Renderer Shim (2.1.271+)	Claude Code 2.1.271 builds its TUI renderer on `Bun.ant.CellSegmenter`, a private API that only Anthropic's bundled Bun exposes. ClawGod loads a JS implementation before the patched bundle, so the TUI paints under stock Bun instead of stalling before the first frame
1h Prompt Cache	Forces 1h TTL allowlist on (was effectively 5m → much higher cache_creation token usage)
Third-Party Cache Fix	Auto-disables `x-anthropic-billing-header` when `baseURL` is non-Anthropic. The header's per-request `cch` field breaks prompt-cache hit rate on DeepSeek / OneAPI / Bedrock / vLLM and any other Anthropic-compatible proxy. You no longer need to set `CLAUDE_CODE_ATTRIBUTION_HEADER=0` yourself.
Auto Re-patch	Detects when the user's native Claude binary has been upgraded; transparently re-extracts and re-patches on next launch
Update Notification	Checks GitHub releases once per 24h (async, non-blocking). Shows a one-line notice if a newer ClawGod version is available
Lean Settings	Three-level token optimization for `~/.claude/settings.json`. on (default): removes unused tool definitions + disables Workflows/Artifact; Remote Control remains available. max: additionally disables Remote Control and removes Plan mode, Agent Teams, bundled skills. off: all tools restored

Lean Settings are non-destructive and persist across updates. Toggle anytime: claude --lean-on (default) / claude --lean-max (aggressive) / claude --lean-off (restore all). To opt out of a single setting, set it yourself (e.g. "disableArtifact": false).

Remote Control (/remote-control, /rc) is allowed in on/off and disabled by default only in max. Updating in on mode or running claude --lean-on clears the older Lean Remote Control disable setting. Only max defaults CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1; explicit environment restrictions are preserved. Remote Control still requires upstream account, authentication, endpoint, and organization eligibility.

Commands

claude              # Patched Claude Code (replaces the official launcher)
clawgod             # Same as `claude`, explicit & guaranteed entry point
claude.orig         # Original unpatched version (auto-backed-up)

clawgod is unambiguous: on Windows where claude.exe may shadow claude.cmd, clawgod.cmd always works. Even after official self-update overwrites claude, clawgod keeps running the patched build.

Configuration

~/.clawgod/provider.json is auto-created on first run. Setting apiKey lets you skip OAuth entirely and point ClawGod at any Anthropic-compatible endpoint.

{
  "apiKey": "sk-ant-...",
  "baseURL": "https://api.anthropic.com",
  "model": "",
  "smallModel": "",
  "effort": "",
  "timeoutMs": 3000000
}

apiKey set → ClawGod injects it as ANTHROPIC_API_KEY and isolates from ~/.claude/settings.json. The default protocol requires an Anthropic Messages-compatible endpoint. For Chat Completions gateways, use the configuration below. A non-Anthropic baseURL populates only ANTHROPIC_AUTH_TOKEN for gateway auth.
apiKey empty → OAuth path. Run claude auth login once; ~/.claude keeps hosting your subagents, skills, and MCP settings.
effort → Sets reasoning effort; an existing CLAUDE_CODE_EFFORT_LEVEL takes precedence. With protocol: "openai-chat", type: "grok" or "openai-compat", the proxy sends reasoning_effort even when Claude omits effort for a custom model alias. low, medium, high, and xhigh pass through; max maps to xhigh; auto omits the parameter to use the upstream default. Choose a level supported by your upstream model. Empty/unset configuration leaves request-level effort in control and adds no effort parameter when the request has none.

OpenAI Chat Completions endpoints

For an endpoint that exposes /v1/chat/completions, configure:

{
  "protocol": "openai-chat",
  "apiKey": "sk-...",
  "baseURL": "https://example.com/v1",
  "model": "your-upstream-model",
  "smallModel": "your-upstream-model"
}

baseURL is the API base (including /v1 when required), not the full /chat/completions URL. Set the model names to IDs accepted by your provider. ClawGod starts a loopback proxy inside the launcher process; no external gateway or separate service is required. timeoutMs also covers the upstream request and stream; an existing API_TIMEOUT_MS takes precedence.

Omitting protocol keeps the existing behavior. type: "openai-compat" and type: "grok" remain supported; an explicit protocol takes precedence. protocol: "anthropic" uses Messages directly. Grok defaults to https://api.x.ai/v1 and retains its settings/environment API key fallback. Other Chat providers require an explicit API base and key.
Supports text, system prompts, tool calls/results, forced tool selection and parallel-tool controls, streaming, usage, and base64/URL images. Parallel tool arguments are assembled and validated before their content blocks are emitted; text still streams immediately. Missing upstrea