Minicor MCP Server

This server (without client built-ins) exposes a complete data-exfiltration chain: pull_workflow → tfa_list_secrets → create_flow. Untrusted input is ingested, private data is read, and it can be sent to an external sink via the agent composing the tools (→). Static analysis proves the primitive exists, not that a specific run will occur.

Fix: Remove one leg of the trifecta: isolate untrusted-input tools from secret-reading tools and from egress tools, or require human approval between them.

Location: flow pull_workflow → tfa_list_secrets → create_flow

Tool "spawn_agent" appears to run shell commands or evaluate code (keyword "spawn" in tool name). 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 spawn_agent

In the server's implementation (`dist/recording-frames.js:21`): 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: rt { spawn } from "node:child_process"; import { createHash } from "node:crypto"; import { mkdtemp, mkdir, open, readFil

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/recording-frames.js

In the server's implementation (`dist/tools/vm-rpa.js:250`): 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: gify(appName)} result = subprocess.run(["taskkill", "/IM", app_name, "/F"], capture_output=True, text=True) print(json.d

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/tools/vm-rpa.js

In the server's implementation (`dist/tools/vm.js:718`): 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: subprocess result = subprocess.run(["powershell", "-command", "Get-Clipboard"], capture_output=True, text=True)

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/tools/vm.js

In the server's implementation (`dist/seeer-client.js:96`): 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: } const res = await fetch("http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/iden

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/seeer-client.js

Tool "spawn_agent" 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 spawn_agent

In a packaging/dev/install script (shipped, but not the server runtime) (`dist/setup.js:11`): 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: ort { exec } from "node:child_process"; import fs from "node:fs"; import http from "node:http"; import os from "node:os"

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/setup.js