菜单
Development conventions and architecture guide for the Clawd Code CLI repository — a Solana-native AI coding agent with perps trading, image generation, voice, and x402 payments.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Register, discover, and hire autonomous AI agents on Solana. On-chain identity via Metaplex Core NFTs (8004-solana SDK or REST API). Search by capability, check reputation backed by $CLAWD payment proofs, browse the service catalog, and get machine-readable hiring instructions. Supports Google A2A agent-card.json and Anthropic MCP server-card.json. No EVM — Solana only.
Build Solana applications with Helius infrastructure. Covers transaction sending (Sender), asset/NFT queries (DAS API), real-time streaming (WebSockets, Laserstream), event pipelines (webhooks), priority fees, wallet analysis, agent onboarding, and handoffs to Clawd agent/x402/pump.fun workflows.
Build, package, run, and monetize Clawd agent surfaces from the clawd-agents packages. Covers CLAWD agent product registration and SSO, clawd-go, Clawd Grok, Solana x402 gateways, A2A/MCP paid agent calls, and pump.fun/PumpSwap bot operations.
Install themed spinner verb packs for Clawd agents. Fetches spinner packs directly from the solana-clawd repository and installs them into the user's settings. Use when the user wants to customize their Clawd spinner, install spinner verbs, or change spinner themes.
Drive the on-chain clawd-zk program (nullifiers, Groth16 proofs, Light Protocol compressed state) from a Clawd agent or REPL. Use when the user wants to attest a model, commit encrypted state, verify a proof, derive a nullifier, or ask a natural-language question that maps to any of those verbs.
Build Solana trading applications combining DFlow trading APIs with Helius infrastructure. Covers spot swaps (imperative and declarative), prediction markets, real-time market streaming, Proof KYC, the DFlow Agent CLI for autonomous trading, transaction submission via Sender, fee optimization, shred-level streaming via LaserStream, and wallet intelligence.
| name | clawd-code-skill |
| description | Development conventions and architecture guide for the Clawd Code CLI repository — a Solana-native AI coding agent with perps trading, image generation, voice, and x402 payments. |
Clawd Code is the Solana-native AI coding CLI for the Clawd ecosystem. It supports file editing, shell commands, git workflows, code review, multi-agent coordination, IDE integration (VS Code, JetBrains), Model Context Protocol (MCP), perpetuals trading on Phoenix DEX, image generation, voice synthesis, and x402 payment-gated commerce.
Codebase: ~1,900 files, 512,000+ lines of TypeScript under src/.
| Component | Technology |
|---|---|
| Language | TypeScript (strict mode, ES modules) |
| Runtime | Bun (JSX support, bun:bundle feature flags) |
| Terminal UI | React + Ink (React for CLI) |
| CLI Parser | Commander.js (@commander-js/extra-typings) |
| API Client | @anthropic-ai/sdk (Claude), native fetch (Grok, DeepSeek, OpenRouter) |
| Validation | Zod v4 |
| Linter/Formatter | Biome |
| Analytics | PostHog (self-hosted analytics) |
| Protocol | Model Context Protocol (MCP) |
| Blockchain | @solana/web3.js, Phoenix SDK, Vulcan MCP |
| Payments | x402 (HTTP 402 Payment Required via Solana) |
src/)| Directory | Purpose |
|---|---|
commands/ | ~50 slash commands (/help, /wallet, /perps, /arena, etc.) |
tools/ | ~40 agent tools (Bash, FileRead, FileWrite, Glob, Grep, etc.) |
components/ | ~140 Ink/React UI components for terminal rendering |
services/ | External integrations (API, OAuth, MCP, LSP, analytics, plugins) |
bridge/ | Bidirectional IDE communication layer |
state/ | React context + custom store (AppState) |
hooks/ | React hooks (permissions, keybindings, commands, settings) |
types/ | TypeScript type definitions |
utils/ | Utilities (shell, file ops, permissions, config, git) |
screens/ | Full-screen UIs (Doctor, REPL, Resume, Compact) |
skills/ | Bundled skills + skill loader system |
plugins/ | Plugin system (marketplace + bundled plugins) |
coordinator/ | Multi-agent coordination & supervisor logic |
tasks/ | Task management (shell tasks, agent tasks, teammates) |
context/ | React context providers (notifications, stats, FPS) |
memdir/ | Persistent memory system (CLAUDE.md, user/project memory) |
entrypoints/ | Initialization logic, Agent SDK, MCP entry |
voice/ | Voice input/output (STT, TTS via sherpa-onnx) |
vim/ | Vim mode keybinding support |
schemas/ | Zod configuration schemas |
keybindings/ | Keybinding configuration & resolver |
migrations/ | Config migrations between versions |
outputStyles/ | Output formatting & theming |
query/ | Query pipeline & processing |
server/ | Server/daemon mode |
remote/ | Remote session handling |
| File | Role |
|---|---|
src/main.tsx | CLI entry point (Commander parser, startup profiling) |
src/QueryEngine.ts | Core LLM API caller (streaming, tool-call loops) |
src/Tool.ts | Tool type definitions & buildTool factory |
src/tools.ts | Tool registry & presets |
src/commands.ts | Command registry |
src/context.ts | System/user context collection (git status, memory) |
src/cost-tracker.ts | Token cost tracking |
src/env.ts | Multi-provider env loader (Grok, Claude, DeepSeek) |
src/arena.ts | Metaplex Agent Arena on-chain identity |
src/main.tsx — Commander CLI parser, startup profilingsrc/entrypoints/init.ts — Config, telemetry, OAuth, provider setupsrc/entrypoints/cli.tsx — CLI session orchestrationsrc/entrypoints/mcp.ts — MCP server modesrc/entrypoints/sdk/ — Agent SDK (programmatic API)src/replLauncher.tsx — REPL session launcherStartup performs parallel initialization: keychain prefetch, feature flag checks, wallet detection, then core init.
Each tool lives in src/tools/{ToolName}/ and uses buildTool:
export const MyTool = buildTool({
name: 'MyTool',
aliases: ['my_tool'],
description: 'What this tool does',
inputSchema: z.object({
param: z.string(),
}),
async call(args, context, canUseTool, parentMessage, onProgress) {
// Execute and return { data: result, newMessages?: [...] }
},
async checkPermissions(input, context) { /* Permission checks */ },
isConcurrencySafe(input) { /* Can run in parallel? */ },
isReadOnly(input) { /* Non-destructive? */ },
prompt(options) { /* System prompt injection */ },
renderToolUseMessage(input, options) { /* UI for invocation */ },
renderToolResultMessage(content, progressMessages, options) { /* UI for result */ },
})
Directory structure per tool: {ToolName}.ts or .tsx (main), UI.tsx (rendering), prompt.ts (system prompt), plus utility files.
Commands live in src/commands/ and follow three types:
const command = {
type: 'prompt',
name: 'my-command',
description: 'What this command does',
progressMessage: 'working...',
allowedTools: ['Bash(git *)', 'FileRead(*)'],
source: 'builtin',
async getPromptForCommand(args, context) {
return [{ type: 'text', text: '...' }]
},
} satisfies Command
Commands are registered in src/commands.ts and invoked via /command-name in the REPL.
Box, Text, useInput())src/components/design-system/AppState via React context + custom store (src/state/AppStateStore.ts)src/state/onChangeAppState.tsdefault (prompt per operation), plan (show plan, ask once), bypassPermissions (auto-approve), auto (ML classifier)Bash(git *), FileEdit(/src/*)checkPermissions() returning { granted: boolean, reason?, prompt? }Bun's bun:bundle feature flags enable dead-code elimination at build time:
import { feature } from 'bun:bundle'
if (feature('PROACTIVE')) { /* proactive agent tools */ }
Notable flags: PROACTIVE, KAIROS, BRIDGE_MODE, VOICE_MODE, COORDINATOR_MODE, DAEMON, WORKFLOW_SCRIPTS.
| Element | Convention | Example |
|---|---|---|
| Files | PascalCase (exports) or kebab-case (commands) | BashTool.tsx, commit-push-pr.ts |
| Components | PascalCase | App.tsx, PromptInput.tsx |
| Types | PascalCase, suffix with Props/State/Context | ToolUseContext |
| Hooks | use prefix | useCanUseTool, useSettings |
| Constants | SCREAMING_SNAKE_CASE | MAX_TOKENS, DEFAULT_TIMEOUT_MS |
.js extensions (Bun convention)const getModule = () => require('./heavy.js')process.envbiome-ignore markers for manual import ordering where needed| Service | Path | Purpose |
|---|---|---|
| API | services/api/ | Multi-provider AI client (Grok/Claude/DeepSeek) |
| MCP | services/mcp/ | MCP client, tool/resource discovery |
| OAuth | services/oauth/ | OAuth 2.0 auth flow |
| LSP | services/lsp/ | Language Server Protocol manager |
| Analytics | services/analytics/ | PostHog analytics, telemetry, events |
| Plugins | services/plugins/ | Plugin loader, marketplace |
| Compact | services/compact/ | Context compression |
| x402 Payments | services/x402/ | HTTP 402 payment routing on Solana |
| Vulcan | services/vulcan/ | Phoenix perps MCP integration |
| Token Estimation | services/tokenEstimation.ts | Token count estimation |
Settings locations:
~/.clawd-code/config.json, ~/.clawd-code/settings.json.clawd/config.json, .clawd/settings.json~/.grok/config.toml and ./.grok/config.toml for xAI Grok-style settings| Variable | Purpose |
|---|---|
ZAI_API_KEY | Z.AI API key for GLM-5.2 / GLM-5V / GLM-Image (default provider) |
ZAI_AGENT_BASE_URL | Z.AI Agent API base URL for slides_glm_agent |
ZAI_CHART_MODEL | Chart/report planning model (glm-5.2) |
ZAI_CHART_VISION_MODEL | Chart mode vision model (glm-5v-turbo) |
ZAI_THINKING | Z.AI thinking mode (enabled or disabled) |
ZAI_REASONING_EFFORT | GLM effort: max, xhigh, high, medium, low, minimal, none |
XAI_API_KEY | xAI Grok API key (optional provider + voice agent) |
ANTHROPIC_API_KEY | Anthropic Claude API key |
DEEPSEEK_API_KEY | DeepSeek API key |
OPENROUTER_API_KEY | OpenRouter API key |
OPENROUTER_NEMO_MODEL1 | Balanced/free OpenRouter Nemo route |
OPENROUTER_NEMO_MODEL2 | Most-intelligent OpenRouter Nemo route |
OPENROUTER_NEMO_MODEL3 | Fast/free OpenRouter Nemo route |
OPENROUTER_FABLE5 | OpenRouter Claude Fable 5 route |
OPENROUTER_FABLE_LATESY | OpenRouter Claude Fable latest route |
HELIUS_API_KEY | Helius RPC/webhook access |
SOLANA_RPC_URL | Custom Solana RPC endpoint |
SOLANA_CLUSTER | Solana harness cluster label |
SOLANA_COMMITMENT | Solana harness commitment |
SOLANA_HARNESS_READONLY | Blocks mutation RPC by default |
LIVE_TRADING | Enable live perps trading (default: false) |
OPERATOR_CONFIRMED | Required operator acknowledgement for live execution |
PERPS_SIM_ONLY | Must be false before real perps execution |
IMPERIAL_ENABLED | Enable Imperial API reads/routing |
IMPERIAL_LIVE | Enable Imperial live order path after global gates pass |
IMPERIAL_WALLET | Wallet pubkey bound to the Imperial JWT |
IMPERIAL_JWT | Imperial mobile JWT; delegated trading credential |
IMPERIAL_PROFILE_INDEX | Isolated Imperial profile index (0..5) |
IMPERIAL_DEFAULT_UNDERWRITER | Imperial venue; 2 is Phoenix |
Clawd Code connects to multiple MCP servers for Solana operations:
| MCP Server | Tools | Purpose |
|---|---|---|
| Helius | DAS queries, webhooks, priority fees | Solana blockchain access |
| Vulcan | Trade execution, portfolio, market data | Phoenix perps trading |
| Imperial | Funding, portfolio, and live order routing | Solana perps router, Phoenix preferred |
| DFlow | Spot swaps, prediction markets | DeFi routing |
| Jupiter | Swap quotes, token lists | DEX aggregation |
bun:bundle feature flags or env checks.checkPermissions().LIVE_TRADING=true, OPERATOR_CONFIRMED=true, and PERPS_SIM_ONLY=false.clawd-code chain for Solana RPC inspection and chain ask for Z.AI-assisted harness planning.IMPERIAL_JWT like a private trading credential; never display, log, or commit it.