Hyperpolymath Boj MCP Server

In the server's implementation (`mcp-bridge/lib/security.js:49`): 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: string only — not a JS eval() call; Hypatia js_eval finding is a false positive "exec(", ]; /** * Normalize a strin

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 mcp-bridge/lib/security.js

In the server's implementation (`mcp-bridge/lib/nickel-validator.js:33`): 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: spawnSync } from "node:child_process"; import { existsSync } from "node:fs"; import { fileURLToPath } from "node:url";

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 mcp-bridge/lib/nickel-validator.js

In the server's implementation (`mcp-bridge/lib/security.js:50`): 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: is a false positive "exec(", ]; /** * Normalize a string for injection analysis. * Strips zero-width characters, n

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 mcp-bridge/lib/security.js

In the server's implementation (`cartridges/docker-hub-mcp/mod.js:278`): 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 response = await fetch("https://registry-1.docker.io/v2/ratelimitpreview/test/manifests/latest", { method

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 cartridges/docker-hub-mcp/mod.js

In the server's implementation (`mcp-bridge/lib/api-clients.js:188`): 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("https://api.github.com/graphql", { method: "POST", headers: { "Author

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 mcp-bridge/lib/api-clients.js

In the server's implementation (`elixir/priv/js_pool_worker.js:39`): 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 imported = await import(modUrl); handleTool = imported.handleTool; modCache.set(mod, handleTool);

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 elixir/priv/js_pool_worker.js

In the server's implementation (`elixir/priv/js_runner.js:42`): 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: y { const mod = await import(modUrl); if (typeof mod.handleTool !== "function") { console.error(JSON.stringify(

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 elixir/priv/js_runner.js