Back to skill

Security audit

ClawdTable

Security checks for vulnerabilities and agentic risk

Overview

This Solana casino skill is mostly coherent with its stated purpose, but it needs Review because joining can automatically move the wallet's full USDC balance into the app vault and several identity, key-storage, and endpoint risks are under-disclosed.

Review carefully before installing. Use only a disposable wallet, keep minimal funds in it, prefer the official secure endpoint, avoid putting real identity information in OpenClaw or CLAWDTABLE_DISPLAY_NAME, and do not run join or play-poker unless you accept that wallet USDC may be moved into the ClawdTable vault automatically.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (3)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
clawdtable-cli.js:613
Finding

Join Command Automatically Deposits the Entire Wallet USDC Balance

Content
View full analysis
0) { console.log(`Depositing ${walletUsdc / 1e6} USDC to vault...`); await prog.methods .deposit(new BN(walletUsdc)) .accountsPartial({ wallet: keypair.publicKey, agentTokenAccount: ata, usdcMint: config.usdcMint, tokenProgram: TOKEN_PROGRAM_ID, }) .rpc(); console.log(`Deposited ${walletUsdc / 1e6} USDC.`); } } catch { /* no USDC in wallet */ } } } catch { /* vault fetch failed — will be created during register */ } ``` ### Technical Analysis The `join` operation automatically transfers the complete USDC balance of the wallet into the ClawdTable vault whenever the vault's available balance is zero. It does not request an amount, obtain a dedicated confirmation, impose a maximum, or preserve a wallet reserve. This exceeds the minimum authority needed to join a table. The documented onboarding flow presents joining and depositing as separate actions and shows an explicit, limited deposit: ```bash clawdtable join 0 clawdtable deposit 10 ``` A command whose apparent purpose is registration and table admission should not implicitly authorize transfer of every USDC token held by the wall ...[truncated 1620 chars]
Remediation
View remediation
` explicitly. 4. Before signing a deposit, display: - Deposit amount - Current wallet balance - Destination vault - Program ID - Network and RPC endpoint 5. Require explicit confirmation for interactive use or an explicit `--yes` option for automation. 6. Support a configurable deposit cap and reject deposits above that limit. 7. Never infer consent to deposit the complete balance merely from consent to join a table. 8. Add tests confirming that `join` and `play poker` cannot change token balances unless an explicit deposit amount was supplied. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
clawdtable-cli.js:57
Finding

OpenClaw Agent Identity Is Read and Transmitted Without Explicit Disclosure or Consent

Content
View full analysis
0 && agents[0].identity?.name) { AGENT_NAME = agents[0].identity.name; } } catch { /* no openclaw config */ } const DISPLAY_NAME = process.env.CLAWDTABLE_DISPLAY_NAME || AGENT_NAME; ``` The derived identity is subsequently sent to the configured WebSocket service: ```js function wsSession(address, name, callback, timeoutMs = 15000) { return new Promise((resolve, reject) => { const ws = new WebSocket(SERVER_URL); let keypair = null; const timer = setTimeout(() => { ws.close(); reject(new Error("Timeout waiting for server response")); }, timeoutMs); ws.on("open", () => { ws.send(JSON.stringify({ type: "identify", address, name })); }); ``` ### Technical Analysis The CLI directly reads `$HOME/.openclaw/openclaw.json`, parses the complete configuration document, extracts the first configured agent's identity name, and uses that value as the default public display name. WebSocket operations then transmit the name together with the wallet address to `CLAWDTABLE_SERVER_URL`. The Skill metadata declares `CLAWDTABLE_SERVER_URL` as an environmental requirement, but it does not disclose that the Skill reads an external OpenClaw configuration file or automatically exports an identity from it. This behavior is not necessary to play the game: a generic pseudonym or an explicitly provided `CLAWDTABLE_DISPLAY_NAME` would be sufficient. Reading the entire configura ...[truncated 1514 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
clawdtable-cli.js:194
Finding

Configurable Plaintext Transport and Unscoped Wallet Challenge Signing

Content
View full analysis
{ const ws = new WebSocket(SERVER_URL); let keypair = null; const timer = setTimeout(() => { ws.close(); reject(new Error("Timeout waiting for server response")); }, timeoutMs); ws.on("open", () => { ws.send(JSON.stringify({ type: "identify", address, name })); }); const messages = []; ws.on("message", async (raw) => { const msg = JSON.parse(raw.toString()); messages.push(msg); if (msg.type === "challenge") { // Sign the challenge with Ed25519 keypair try { if (!keypair) keypair = loadKeypair(); const messageBytes = Buffer.from(msg.challenge, "utf-8"); const signature = nacl.sign.detached(messageBytes, keypair.secretKey); ws.send(JSON.stringify({ type: "challengeResponse", signature: Buffer.from(signature).toString("base64"), })); ``` HTTP helpers also deliberately permit plaintext transport: ```js function fetchJson(url) { return new Promise((resolve) => { const lib = url.startsWith("https") ? require("https") : require("http"); lib.get(url, { timeout: 5000 }, (res) => { let data = ""; res.on("data", (chunk) => { data += chunk; }); res.on("end", () => { try { resolve(JSON.parse(data)); } catch { resolve(null); } }); }).on("error", () => resolve(null)); }); } ``` ### Technical Analysis ...[truncated 3054 chars]
Remediation
View remediation
Nonce: Purpose: websocket-authentication Issued-At: Expires-At: ``` 6. Validate the protocol identifier, origin, wallet, nonce length, timestamp, purpose, and short expiration before signing. 7. Track recently signed nonces locally for the session and reject duplicate challenges. 8. Avoid using the wallet key for generic authentication where possible; use a dedicated session authentication key bound to the wallet through a narrowly scoped authorization. 9. Prefer direct Solana RPC submission for all signed transactions rather than sending them through an application-controlled relay. 10. Clearly display the target network, program, accounts, instruction, and amount before signing any transaction with financial effects. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (20)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description centers on blackjack, but the file also advertises poker, wallet/vault fund transfers, balance inspection, local keypair management, and referenced funding behavior. That mismatch is dangerous because users and orchestrators may grant trust based on a narrower purpose while the skill actually includes broader financial and sensitive-key capabilities that can expose funds or secrets.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill metadata advertises blackjack, but the CLI contains undisclosed poker gameplay, room selection, and poker transaction submission logic. This hidden capability expands the asset-handling and gambling surface beyond what a user or platform reviewer would reasonably expect, undermining informed consent and increasing the chance that an agent is induced to perform unintended financial actions.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The poker path both introduces an undeclared game mode and performs automatic faucet requests and vault deposits. Because poker is outside the declared purpose, this creates a particularly misleading financial-action path where an agent may acquire funds and move USDC in support of functionality the user was never told existed.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill metadata and description frame this as a blackjack skill, but the IDL exposes a much broader attack surface including poker table creation, joins, betting, dealing, showdown, and admin/test shuffle operations. This is dangerous because agents or integrators may grant trust, permissions, or funds based on the narrower blackjack description while unknowingly interacting with undisclosed gambling and fund-handling functionality.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill exposes environment and network-dependent behavior but does not declare an explicit tool scope or permission boundary. In a financial skill that can create wallets, connect to remote infrastructure, and initiate blockchain actions, missing scope metadata makes it harder for hosts and users to reason about what the skill is allowed to access and increases the chance of overbroad execution.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The quick-start and command sections encourage immediate use of commands that can create wallets, deposit funds, withdraw funds, and place bets, but they do not clearly warn that real USDC and on-chain transactions are at risk. In a gambling/crypto context, this omission materially raises the risk of accidental fund movement or loss because users may treat the skill like a harmless game integration rather than a real-money financial tool.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The code generates and persists the full Ed25519 secret key to a predictable path on disk in plaintext JSON, with no encryption, no passphrase, and no user-facing disclosure. Local malware, other users on a shared host, backups, or accidental file exposure could compromise the private key and allow unauthorized signing of transactions.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The join flow automatically contacts a remote faucet and later auto-deposits available USDC without explicit confirmation, which exceeds the minimal expectations of a game/chat skill. Even if intended for convenience, automatic funding and asset movement are sensitive operations because they can alter wallet state and commit funds based on server-driven setup rather than an explicit user decision at the moment of transfer.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The join flow automatically deposits any wallet USDC into the program vault when the vault is empty, without explicit confirmation. This is dangerous because it moves assets from a user-controlled token account into application custody as a side effect of joining, which can surprise users and increase exposure to program or business-logic risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The poker setup path also auto-deposits wallet USDC into the vault without explicit confirmation. Because this occurs in an undisclosed poker feature, the consent problem is amplified: funds may be moved in support of a hidden capability the user did not agree to use.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest advertises that agents can 'chat with other agents,' but the published interface contains no messaging or chat primitives, only game/account operations and lounge presence tracking. This mismatch is risky because it misleads users about what actions the skill can actually perform and can hide the fact that the skill is primarily a real-money transaction and gambling interface rather than a communication tool.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill explicitly supports deposit, betting, and withdrawal of USDC on Solana, yet the user-facing description lacks a clear warning that invoking the skill can move real assets and cause financial loss. In the context of an AI-agent casino skill, this omission is especially dangerous because an agent or operator may treat actions as game-like or simulated when they are actually authorizing irreversible token transfers.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill includes chat features but does not prominently warn, at the point of feature introduction, that messages are visible to other agents and spectators and may be retained server-side. In an adversarial multi-agent environment, users may disclose strategies, wallet-related details, or other sensitive information that can be harvested or used for manipulation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The join flow automatically POSTs the user's wallet address and requested amount to a server faucet endpoint when the balance is low. Although it logs that funds are being requested, it does not clearly disclose that wallet information is being sent to a remote service as part of that step.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
91% confidence
Finding

The dependency uses a caret range instead of an exact version, so installs may resolve to different releases over time, including versions with newly disclosed vulnerabilities or breaking behavior. In a crypto-gambling skill that handles wallets, network traffic, and potentially funds, supply-chain drift materially increases security risk because a later compatible release could introduce exploitable code or pull in a compromised package.

Content

Scanner excerpt · package.json (reported line 6)May include surrounding context.

json
"version": "1.0.2",
  "description": "ClawdTable blackjack CLI for AI agents on Solana",
  "dependencies": {
    "ws": "^8.19.0",
    "@solana/web3.js": "^1.98.0",
    "@coral-xyz/anchor": "^0.32.1",
    "tweetnacl": "^1.0.3"

Unverifiable Dependency: ws has 7 known advisory(ies) (CVE-2016-10518 (Remote Memory Disclosure in ws); CVE-2024-37890 (ws affected by a DoS when handling a request with many HTTP headers); CVE-2026-45736 (ws: Uninitialized memory disclosure) +4 more), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
85% confidence
Finding

The manifest does not pin ws, so it is impossible to verify whether the installed version avoids known vulnerabilities, including denial-of-service and memory disclosure issues. Because this skill likely maintains live WebSocket connections for gameplay and chat, an affected ws version could expose availability or data confidentiality problems in a network-facing component.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
96% confidence
Finding

Using an unpinned version of @solana/web3.js is especially risky because this package directly interacts with Solana wallets, transactions, and key material. A semver-range install can silently pull a vulnerable or maliciously republished release, which is more dangerous here given the skill's financial context and the package's history of advisories related to crashes and malware.

Content

Scanner excerpt · package.json (reported line 7)May include surrounding context.

json
"description": "ClawdTable blackjack CLI for AI agents on Solana",
  "dependencies": {
    "ws": "^8.19.0",
    "@solana/web3.js": "^1.98.0",
    "@coral-xyz/anchor": "^0.32.1",
    "tweetnacl": "^1.0.3"
  }

Unverifiable Dependency: @solana/web3.js has 3 known advisory(ies) (CVE-2024-30253 (Handling untrusted input can result in a crash, leading to loss of availability ); CVE-2024-54134 (Modified package published to npm, containing malware that exfiltrates private k); MAL-2024-11183 (Malicious code in @solana/web3.js (npm))), but the manifest does not pin a version, so it is unknown whether the installed release is affected

Low
Category
Supply Chain
Confidence
94% confidence
Finding

The unpinned @solana/web3.js dependency cannot be verified against known advisories, including reports of malicious package publication and vulnerabilities affecting availability or private key exposure. In a Solana-based gambling skill that may manage wallets and transactions, uncertainty about the exact installed version creates a significant supply-chain risk with direct financial consequences.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
88% confidence
Finding

The Anchor dependency is specified with a caret range, allowing automatic uptake of future releases without explicit review. For software that signs or submits blockchain transactions, dependency drift can change serialization, transaction logic, or introduce vulnerable code paths that affect correctness and security.

Content

Scanner excerpt · package.json (reported line 8)May include surrounding context.

json
"dependencies": {
    "ws": "^8.19.0",
    "@solana/web3.js": "^1.98.0",
    "@coral-xyz/anchor": "^0.32.1",
    "tweetnacl": "^1.0.3"
  }
}

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
90% confidence
Finding

tweetnacl is a cryptographic library, so version drift is more sensitive than for ordinary utilities because subtle changes can affect key handling, signatures, or randomness assumptions. In a casino skill using USDC on Solana, compromise or regression in a crypto primitive dependency could directly affect authentication or transaction security.

Content

Scanner excerpt · package.json (reported line 9)May include surrounding context.

json
"ws": "^8.19.0",
    "@solana/web3.js": "^1.98.0",
    "@coral-xyz/anchor": "^0.32.1",
    "tweetnacl": "^1.0.3"
  }
}