Gas Fakes MCP Server

In the server's implementation (`src/cli/executor.js:200`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate 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: const gasFunction = new Function(mainScript); res = await gasFunction(); } if (res) { const output = typ

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 src/cli/executor.js

In the server's implementation (`src/services/libhandlerapp/fakelibrary.js:101`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate 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: his[this.userSymbol] = (new Function(this.wrapper))() return this } get manifest() { if (!this.__manifest)

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 src/services/libhandlerapp/fakelibrary.js

The implementation both builds shell commands out of runtime values (`src/services/jdbc/fakejdbcservice.js`) and evaluates runtime values as code (`src/cli/executor.js`). Each is a separate arbitrary-execution primitive; a server exposing both gives anything that reaches either one a direct path to running attacker-chosen code. These two sinks are in different files — confirm whether they are actually connected, or are unrelated code paths (e.g. a vendored bundle plus a CLI wrapper).

Fix: Remove the dynamic eval, and pass command arguments as an argv array instead of building a shell string. If both are genuinely required, constrain and validate every value that can reach them.

Location: server implementation

Tool "run_script" appears to run shell commands or evaluate code (keyword "run_script" in tool name, parameter "script"). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.

Fix: Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.

Location: tool run_script

In the server's implementation (`src/cli/executor.js:197`): 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: const gasFunction = new Function("args", mainScript); res = await gasFunction(args); } else { const gasFun

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 src/cli/executor.js

In the server's implementation (`src/services/html/consumerworker.js:95`): 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: ry { result = new Function('_he', ...scopeKeys, code)(htmlEscape, ...scopeVals); } catch (e) {

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 src/services/html/consumerworker.js

In the server's implementation (`src/services/libhandlerapp/fakelibrary.js:101`): 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: his[this.userSymbol] = (new Function(this.wrapper))() return this } get manifest() { if (!this.__manifest)

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 src/services/libhandlerapp/fakelibrary.js

In the server's implementation (`src/cli/mcp.js:4`): 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: import { spawn } from "child_process"; import fs from "fs"; import os from "os"; import path from "path"; import { file

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 src/cli/mcp.js

In the server's implementation (`src/services/jdbc/fakejdbcservice.js:3`): 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: port { execSync } from 'child_process'; class FakeJdbcService { constructor() { this.__fakeObjectType = 'Jdbc';

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 src/services/jdbc/fakejdbcservice.js

In the server's implementation (`src/support/msgraph/msauth.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 { readFile, writeFile } from 'fs/promises'; import { existsSync } from 'f

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 src/support/msgraph/msauth.js

In the server's implementation (`lagos/create_top_owners_sheet.js:13`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: response = UrlFetchApp.fetch(`https://www.googleapis.com/oauth2/v3/tokeninfo?access_token=${accessToken}`) if (respon

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 lagos/create_top_owners_sheet.js

In the server's implementation (`src/support/auth.js:256`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: t tokenResponse = await fetch("https://oauth2.googleapis.com/token", { method: 'POST', headers: { 'C

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 src/support/auth.js

In the server's implementation (`src/services/jdbc/fakejdbcservice.js:86`): 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(`pgrep -f "cloud.*sql.*proxy.*${instanceName}"`, { stdio: "ignore", }

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 src/services/jdbc/fakejdbcservice.js

Tool "run_script" takes a command-shaped parameter "script" with no enum/pattern constraint. Free-form, model- or attacker-controlled arguments reaching a shell is the command-injection precondition.

Fix: Constrain the parameter (enum/pattern), or build the command from a fixed template with escaped args.

Location: tool run_script · inputSchema.properties.script

In the server's implementation (`src/services/html/consumerworker.js:36`): 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: onst userModule = await import(mainScriptPath); // Expose all exports to globalThis for legacy patterns Obj

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 src/services/html/consumerworker.js

In the server's implementation (`src/support/metadata.js:12`): 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 = {}; const pjson = require(packageJsonPath); if (!globalThis.GasFakes.metadata) globalThis.GasFakes.metadata = {

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 src/support/metadata.js

Tool "run_script" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool run_script

In a packaging/dev/install script (shipped, but not the server runtime) (`src/cli/setup.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: port { execSync } from "child_process"; import { checkForGcloudCli, runCommandSync, runCommandWithRetrySync } from "./ut

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 src/cli/setup.js