Agentchannel MCP Server

In the server's implementation (`dist/cli.js:571`): 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("node:child_process").then(({ exec }) => { exec(`open http://localhost:${opts.port}`);

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.js

In the server's implementation (`dist/mqtt-client.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: { execSync } from "node:child_process"; import { deriveKey, deriveSubKey, hashRoom, hashSub, deriveDmKey, hashDm, encryp

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/mqtt-client.js

Untrusted-input tools ([get_message, read_messages]) co-exist with external-action tools ([create_webhook, delete_hook, send_message]). A prompt injection could cause unwanted external actions, though no direct sensitive-data leak path was found.

Evidence: untrusted [get_message, read_messages] → sinks [create_webhook, delete_hook, send_message]

Fix: Require confirmation for state-changing/egress actions triggered after processing untrusted content.

Location: flow get_message → create_webhook

In the server's implementation (`dist/cli.js:165`): 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.agentchannel.workers.dev/invites?token=${opts.token}`); const dat

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.js

In the server's implementation (`dist/server.js:34`): 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://registry.npmjs.org/agentchannel/latest", { signal: AbortSignal.timeout(3000) });

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.js

In the server's implementation (`dist/tools/create.js:31`): 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.agentchannel.workers.dev/invites", { method: "POST",

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/create.js

In the server's implementation (`dist/tools/registry.js:27`): 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.agentchannel.workers.dev/invites", { method: "POST",

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/registry.js

In the server's implementation (`dist/update-check.js:49`): 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://registry.npmjs.org/agentchannel/latest", { signal: AbortSignal.ti

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/update-check.js

In the server's implementation (`ui/app.js:713`): 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: ); if (!ch) return; fetch("https://api.agentchannel.workers.dev/invites", { method: "POST", headers: {"Conte

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 ui/app.js

In the server's implementation (`dist/cli.js:572`): 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: xec }) => { exec(`open http://localhost:${opts.port}`); }); } }); // ── send ───────────────────

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.js

Tool "create_webhook" takes a URL/host parameter "url" with no allowlist/pattern. An outbound-request tool with an unbounded destination enables SSRF and cloud-metadata access (e.g. 169.254.169.254).

Fix: Allowlist destinations or constrain the parameter; block private/link-local addresses server-side.

Location: tool create_webhook · inputSchema.properties.url

In a packaging/dev/install script (shipped, but not the server runtime) (`ui/mqtt.min.js:3`): 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: s.regeneratorRuntime=$o:Function("r","regeneratorRuntime = r")($o)}});var _g=j((Ho,mg)=>{A();x();I();(function(t,e){type

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 ui/mqtt.min.js