Is safari-mcp safe?
- Node.js shell/command execution
- npm install-time lifecycle hook
- Dangerous MCP tool capability
What to do: The Swift code was not checked for shell, network or file access, so this result is incomplete. Review that code before installing.
safari-mcp is an npm package analyzed by SkillTotal's deterministic static scanner. The scan found no malicious indicators. Its Swift code was not checked for shell, network or file access, so this result is partial. It can: delegated authentication, install time execution, mcp tools detected, network egress and shell execution — capabilities are what the code can do, not a verdict on intent. Risk score 0/100 (low).
safari-mcp 2.23.0
Get notified when this component changes
Change alerts aren't live yet. Leave your email and we'll let you know when they launch. We use it only for this and delete it on request.
Automated static-analysis result. It can contain false positives and false negatives, and is not a claim about the intent of safari-mcp's authors. Report a false positive.
Behavioral traits
How this component maps to the CSA agentic threat model. Descriptive — it never affects the risk score.
Findings (6)
package.json runs scripts automatically when the package is installed.
"postinstall": "node scripts/postinstall.cjs || true"
Why it matters: Install scripts are a favorite supply-chain foothold — they execute on every machine that installs the package.
Fix: Inspect the hook command. Install-time scripts are a common supply chain execution vector; ensure they do nothing beyond a documented build step.
An MCP tool exposes a powerful capability (files, shell, network, browser, or credentials).
server.tool( "safari_navigate",
server.tool( "safari_navigate_and_read",
server.tool( "safari_screenshot",
server.tool( "safari_screenshot_element",
Why it matters: Wired into an agent, these grant it real access to your machine — confirm each is required.
Fix: Confirm each powerful tool is required and constrained; broad MCP tools (shell/filesystem/network) grant an agent significant host access.
The component can run operating-system commands or spawn processes.
import { execFile, execFileSync } from "node:child_process";import { execFileSync } from "node:child_process";import { execFile, spawn, spawnSync } from "node:child_process";_helperProc = spawn(helperPath, [], { stdio: ["pipe", "pipe", "ignore"] });const proc = spawn("pbcopy", [], { stdio: ["pipe", "ignore", "ignore"] });spawnSync("pbcopy", [], { input: content });const { execFileSync, spawnSync } = require("child_process");const result = spawnSync("codesign", ["-d", "--verbose=2", "--", helper], { encoding: "utf8" });Why it matters: Powerful and often legitimate — confirm the commands aren't built from untrusted input.
Fix: Confirm the command and its arguments are fully controlled and not derived from untrusted input; prefer execFile with an argument array.
The component makes outbound network requests.
_bridgeAuthTokenPromise = fetch(browser.runtime.getURL("bridge-auth-token"), {const response = await fetch(url, { ...init, headers });const response = await fetch(`${_mcpContentWakeBridgeUrl}/content-wakeup`, {import { createServer } from "node:http";const res = await fetch(`http://127.0.0.1:${HTTP_PORT}/proxy-check?profile=${profile}`, {const res = await fetch(`http://127.0.0.1:${HTTP_PORT}/proxy-command`, {// Patch fetch (once)
import { createServer } from "node:http";Why it matters: Usually legitimate, but confirm the destinations are expected and no sensitive data leaves.
Fix: Confirm the destination hosts are expected and that no sensitive data is sent off-host.
An OAuth 2.0 / OpenID Connect delegated-authentication flow was detected (authorization-code / refresh-token / token-exchange grant, an OIDC authorize/discovery endpoint or id_token, or a delegation library). Tools authenticate with the end user's delegated, scoped credentials rather than a long-lived embedded service credential.
!/[?&#](code|token|access_token|id_token|state|session_state)=/i.test(payload.url || "");
Fix: Delegated auth is a lower-blast-radius execution context than an embedded static credential. Confirm the requested scopes are minimal and that tokens are never logged or forwarded off-host.
An MCP tool surface (manifest or tool definitions) was found.
const server = new McpServer({server.tool( "safari_navigate",
server.tool( "safari_go_back",
server.tool( "safari_go_forward",
server.tool( "safari_reload",
server.tool( "safari_read_page",
server.tool( "safari_get_source",
server.tool( "safari_snapshot",
server.tool( "safari_navigate_and_read",
server.tool( "safari_click",
server.tool( "safari_click_and_read",
server.tool( "safari_double_click",
server.tool( "safari_right_click",
server.tool( "safari_native_click",
server.tool( "safari_native_hover",
server.tool( "safari_native_keyboard",
server.tool( "safari_native_type",
server.tool( "safari_fill",
server.tool( "safari_clear_field",
server.tool( "safari_verify_state",
server.tool( "safari_select_option",
server.tool( "safari_react_select_set",
server.tool( "safari_react_select_list_options",
server.tool( "safari_fill_form",
Why it matters: Just context — review which tools it offers and their permissions.
Fix: Review the declared MCP tools and their permissions.
How attackers abuse these capabilities
Interactive labs on the attack class behind the rules above. They show the technique, not anything found in safari-mcp.
Maintain safari-mcp?
The free Claude Code plugin scans each npm, PyPI or MCP package with this report's open-source engine before the install command runs. It runs on your machine and needs no account.
Check your own component
Run the same evidence-backed scan on any MCP server, agent skill, or package.
Scan your own componentHow we determine this: deterministic static analysis (regex + AST), evidence-anchored, no code execution. Methodology →