Vurb Core MCP Server

In the server's implementation (`dist/sandbox/SandboxGuard.js:48`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: unction\s*\(/, reason: 'new Function() has no effect in the sandbox — use direct expressions instead.' }, ]; /** * The

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/sandbox/SandboxGuard.js

In the server's implementation (`dist/cli/commands/create.js:7`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: { execSync } from 'node:child_process'; import { createInterface } from 'node:readline'; import { ProgressTracker } from

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/commands/create.js

In the server's implementation (`dist/cli/commands/deploy.js:183`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: 'SandboxEngine requires child_process and fs, which are unavailable in V8 isolates.')}\n` + ` ${ansi.di

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/commands/deploy.js

In the server's implementation (`dist/cli/commands/introspect.js:43`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: } = await import('node:child_process'); try { execSync('npm install -D esbuild --legacy-peer-deps', { cwd:

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/commands/introspect.js

In the server's implementation (`dist/cli/commands/update.js:39`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: } = await import('node:child_process'); const npmFlags = '--prefer-online --no-fund --no-audit'; // Try batch i

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/commands/update.js

In the server's implementation (`dist/edge-stub.js:92`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: xec = () => CRASH('node:child_process.exec'); export const execSync = () => CRASH('node:child_process.execSync'); export

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/edge-stub.js

In the server's implementation (`dist/introspection/EntitlementScanner.js:34`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: bprocess', identifier: 'child_process', regex: /(?:require\s*\(\s*['"]|import\s*\(\s*['"]|from\s+['"])(?:node:)?child_pr

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/introspection/EntitlementScanner.js

In the server's implementation (`dist/cli/commands/update.js:45`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: try { execSync(`npm install ${npmFlags} ${installArgs.join(' ')}`, { cwd,

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/commands/update.js

In the server's implementation (`dist/prompt/PromptMessage.js:150`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ages.push(PromptMessage.system(`<domain_rules>\n${rules.map(r => `- ${r}`).join('\n')}\n</domain_rules>`)); }

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/prompt/PromptMessage.js

In the server's implementation (`dist/cli/registry.js:34`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: } const mod = await import(fileUrl); /** Extract prompt registry from a module-like object */ function extra

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/registry.js

In the server's implementation (`dist/cli/vurb.js:90`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: { commandYaml } = await import(pkg); await commandYaml(); } catch {

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli/vurb.js

In the server's implementation (`dist/prompt/CursorCodec.js:19`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: nst nodeCrypto = (await import(mod)).webcrypto; if (nodeCrypto != null) { _resolvedCrypto = nodeCryp

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/prompt/CursorCodec.js

In the server's implementation (`dist/server/autoDiscover.js:134`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: s mod = require(filePath); } else { const fileUrl = pathToFileUR

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/server/autoDiscover.js