Back to skill

Security audit

clawCommunity

Security checks for vulnerabilities and agentic risk

Overview

The skill is a real game bridge, but it exposes unauthenticated control and state APIs that can let other callers inspect or command connected game clients.

Install only if you are comfortable running a local game-control bridge and autonomous bot on a trusted machine and network. Do not leave the bridge running unnecessarily; prefer a version that binds WebSocket to loopback, adds authentication, restricts CORS, validates command types, upgrades ws, and provides clear controls to disable or delete persistent memory.

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 (2)

T05 · Unauthorized Access and Privilege Escalation

Error
Location
scripts/OpenClawGameBridge.js:34
Finding

Unauthenticated Network-Accessible WebSocket Control Plane

Content
View full analysis
{ try { const msg = JSON.parse(data); console.log('[Bridge] 收到消息:', msg.type, JSON.stringify(msg).substring(0, 200)); this.handleMessage(msg, ws); } catch (error) { console.error('[Bridge] 消息解析错误:', error); } }); ``` ```javascript // 发送控制指令到游戏客户端 onBridgeSendCommand(ws, msg) { const playerUid = msg.playerUid; const command = msg.command; if (!playerUid || !command) { console.log(`[Bridge] sendCommand 缺少参数: playerUid=${playerUid}, command=`, command); return; } const targetWs = this.clients.get(playerUid); if (!targetWs || targetWs.readyState !== WebSocket.OPEN) { console.log(`[Bridge] sendCommand 玩家不在线: ${playerUid}`); return; } // 包装成 ai_control 消息发给游戏客户端 targetWs.send(JSON.stringify({ type: 'ai_control', playerUid: playerUid, command: command })); console.log(`[Bridge] ✅ 发送指令到 ${playerUid}:`, JSON.stringify(command)); } ``` ```javascript onClientRegister(msg, ws) { // 处理空的playerUid let playerUid = msg.playerUid; if (!playerUid || playerUid === '') { playerUid = 'player_001'; console.log(`[Bridge] ⚠️ 玩家ID为空,已自动设置为: player_001`); } this.clients.set(playerUid, ws); console.log(`[Bridge] ⭐ 客户端注册: playerUid=${playerUid}, ws=${ws.readyState}`); console.log(`[Bridge] 当前保存的客户端:`, Array.from(this.clients.keys())); ``` ### Technical Analysis The WebSocket server is created wit ...[truncated 3120 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/OpenClawGameBridge.js:1449
Finding

Unauthenticated Local HTTP Control API Exposed to Arbitrary Browser Origins

Content
View full analysis
{ // CORS 预检 if (req.method === 'OPTIONS') { res.writeHead(204, { 'Access-Control-Allow-Origin': '*', 'Access-Control-Allow-Methods': 'GET, POST, OPTIONS', 'Access-Control-Allow-Headers': 'Content-Type' }); res.end(); return; } // 强制 JSON 输出 res.setHeader('Content-Type', 'application/json; charset=utf-8'); res.setHeader('Access-Control-Allow-Origin', '*'); const url = req.url.split('?')[0]; const self = this; try { if (req.method === 'GET' && url === '/health') { this.httpHealth(res); } else if (req.method === 'GET' && url === '/clients') { this.httpListClients(res); } else if (req.method === 'GET' && url.startsWith('/clients/')) { const parts = url.split('/'); if (parts[3] === 'perception') { this.httpGetPerception(parts[2], res); } else if (parts[3] === 'mapInfo') { this.httpGetMapInfo(parts[2], res); } else { this.httpNotFound(res, url); } } else if (req.method === 'GET' && url.startsWith('/mapId/')) { const mapId = parseInt(url.split('/')[2]); if (!isNaN(mapId)) { this.httpGetTransportPoints(mapId, res); } else { this.httpError(res, 400, 'Invalid mapId'); } ...[truncated 5900 chars]
Remediation
View remediation
` for every endpoint other than an optional minimal health check. - Store the token with user-only filesystem permissions and never place it in URLs or logs. 2. **Apply player-level authorization** - Associate authenticated controller sessions with permitted player identifiers. - Reject attempts to read or control players outside the caller's authorization scope. 3. **Remove wildcard CORS** - If browser access is unnecessary, omit all CORS response headers. - If browser access is required, allow only the exact trusted game origin, such as the expected HTTPS origin. - Validate the request's `Origin` before returning CORS headers. 4. **Harden command processing** - Allowlist supported command types such as `move`, `sendDialogue`, `interact`, and approved perception requests. - Validate numeric ranges, identifier formats, and maximum message lengths. - Reject unknown fields and malformed command structures. 5. **Limit request bodies** ```javascript let body = ''; const maxBody = 16 * 1024; req.on('data', chunk => { body += chunk; if (Buffer.byteLength(body, 'utf8') > maxBody) { res.writeHead(413); res.end(); req.destroy(); } }); ``` 6. **Reduce exposed information** - Keep `/health` limited to status and uptime. - Do not expose player identifiers in health responses. - Return only the minimum perception fields required by the authenticated caller. 7. **Add abuse controls** - Rate-limit commands and perception requests. - Log authenticated caller identity and command outcomes. - Return generic errors rather than unnecessary internal details. ]]>
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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (47)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The documentation exposes an HTTP/WebSocket bridge with control-command APIs and state-query endpoints while marketing the skill as a simple social-space opener. Exposing control and introspection surfaces under a misleading description makes accidental overtrust more likely and could enable unauthorized client manipulation on the local machine.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 668)May include surrounding context.

md
| `scripts/ai_loop_ws.js` | AI 循环主体(WS 事件驱动 + HTTP 轮询) |

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 669)May include surrounding context.

md
| `scripts/ai_launcher.js` | 交互式启动器 |

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The HTTP /command endpoint forwards arbitrary command objects to a target player with no authentication, origin validation, or command restrictions. Any local process—or potentially a malicious webpage via the permissive CORS policy if a local service is reachable—can issue gameplay control actions to connected clients.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file materially exceeds the declared skill purpose of opening/connecting a bridge and instead runs a persistent autonomous game/social bot with movement, dialogue, goal execution, and memory consolidation. This kind of scope mismatch is dangerous because users and reviewers may authorize the skill expecting a launcher, while it actually performs ongoing autonomous actions and interaction inside the game environment.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill autonomously decides when to move, explore maps, greet players, and send dialogue over a local control channel, effectively acting as a bot. In context, this is more dangerous because the manifest presents the skill as a launcher/social-space opener, so the autonomous command execution is unexpected and can result in unauthorized gameplay actions, spammy interactions, or abuse of the connected client account.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: ws==8.20.0 — 2 advisory(ies): CVE-2026-45736 (ws: Uninitialized memory disclosure); CVE-2026-48779 (ws: Memory exhaustion DoS from tiny fragments and data chunks)

High
Category
Supply Chain
Confidence
97% confidence
Finding

The lockfile pins the project to ws 8.20.0, and the supplied advisory data identifies two known HIGH-severity issues in that version: uninitialized memory disclosure and memory-exhaustion denial of service. Because this skill description indicates an automatically launched bridge component and ws is a WebSocket library commonly exposed to network input, the vulnerable dependency is plausibly reachable and increases risk of remote information disclosure or service disruption.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: ws==8.20.0 — 2 advisory(ies): CVE-2026-45736 (ws: Uninitialized memory disclosure); CVE-2026-48779 (ws: Memory exhaustion DoS from tiny fragments and data chunks)

High
Category
Supply Chain
Confidence
98% confidence
Finding

The package depends on ws 8.20.0, which is flagged with advisories for memory disclosure and memory-exhaustion denial of service. Because this skill explicitly runs a WebSocket bridge between the game client and another component, the vulnerable library is directly exposed in a network-facing role, making exploitation materially more plausible and increasing availability and potential data exposure risk.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
83% confidence
Finding

The skill documents significant capabilities including environment access, local process launch instructions, local HTTP/WebSocket control channels, and persistent storage, but does not declare any tool scope or permissions boundaries. That makes the skill harder to audit and increases the chance an agent may invoke more capability than a user expects, especially given the broad automation claims.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest claims startup is automatic, but the instructions require manual bridge startup and only partial automation afterward. This deceptive or inaccurate setup description can cause users and reviewers to misunderstand what actions the skill will actually perform and what privileges it requires.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation contradicts itself about whether the bridge starts automatically or must be started manually. Contradictory instructions are a security issue here because they obscure actual execution boundaries and can hide where privileged local commands or operator actions are really required.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The activation phrase '打开 clawSpace'/'Open clawSpace' is generic enough to overlap with ordinary conversation, especially because the skill claims broad automatic behavior once triggered. In context, accidental activation could launch a browser, connect to local services, or start an automation workflow unexpectedly.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 76)May include surrounding context.

md
| GET | `/clients/:uid/perception` | 查询感知缓存 | `curl http://localhost:18766/clients/2016386a4_10001/perception` |
| GET | `/clients/:uid/mapInfo` | 查询地图信息 | `curl http://localhost:18766/clients/2016386a4_10001/mapInfo` |
| GET | `/mapId/:mapId/transportPoints` | 查询传送门缓存 | `curl http://localhost:18766/mapId/2000/transportPoints` |
| POST | `/command` | 发送控制指令 | `curl -X POST http://localhost:18766/command -d '{"playerUid":"...","command":{"type":"move","x":25,"y":30}}'` |
| POST | `/perception/request` | 请求感知推送 | `curl -X POST http://localhost:18766/perception/request -d '{"playerUid":"...","category":"npcs"}'` |

**发送控制指令示例 (PowerShell):**

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 77)May include surrounding context.

| GET | /clients/:uid/mapInfo | 查询地图信息 | curl http://localhost:18766/clients/2016386a4_10001/mapInfo | | GET | /mapId/:mapId/transportPoints | 查询传送门缓存 | curl http://localhost:18766/mapId/2000/transportPoints | | POST | /command | 发送控制指令 | curl -X POST http://localhost:18766/command -d '{"playerUid":"...","command":{"type":"move","x":25,"y":30}}' | | POST | /perception/request | 请求感知推送 | curl -X POST http://localhost:18766/perception/request -d '{"playerUid":"...","category":"npcs"}' |

发送控制指令示例 (PowerShell):

powershell

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The stated purpose is a social space, but the documented goals include exploration, monster hunting, and autonomous navigation. That scope expansion makes the skill more dangerous in context because it moves from communication tooling into active gameplay automation and remote control.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill includes persistent memory, event logs, summaries, and long-term curation of encounters and conversations, but this collection is not justified by the lightweight stated purpose. In context, this creates unnecessary privacy risk by storing detailed behavioral and social data beyond what is needed to launch or join a shared space.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill omits a clear upfront warning that it persistently stores local memory, identities, conversations, and also performs retention/deletion management. Without conspicuous disclosure, users cannot provide informed consent for the privacy impact of these features.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The long-term memory design stores other players' identifiers, encounter history, greetings, and conversation content in plain language. In this context, that is sensitive behavioral data about third parties and creates privacy, consent, and possible profiling concerns if the local files are accessed or reused.

Content

No source excerpt is available for this finding.