registry  /  wigolo  /  0.1.40-beta.2

wigolo@0.1.40-beta.2

Local-first web intelligence MCP server for AI coding agents

AI Security Review

scanned 7d ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. No confirmed malicious behavior or install-time hijack was found. The real risk is explicit user-command setup that modifies AI-agent configuration and instructions to register wigolo as an MCP server and prefer its web tools.

Static reason
High-risk behavior combination matched malicious policy.
Trigger
User runs `wigolo setup mcp` or `wigolo init` and selects agents.
Impact
Can change agent web-tool routing and install package-owned instructions/skills; no credential theft, exfiltration, or remote payload execution confirmed.
Mechanism
First-party AI-agent MCP and instruction setup
Rationale
Static inspection supports a warning for explicit first-party AI-agent extension setup, not a publish block. Scanner hits for shell, network, SSRF metadata, and dynamic import map to package-aligned CLI/web-search/plugin features with guards or user invocation.
Evidence
package.jsondist/index.jsdist/cli/setup-mcp.jsdist/cli/init.jsdist/cli/tui/config-writer.jsdist/cli/agents/codex.jsdist/cli/agents/claude-code.jsdist/cli/agents/utils.jsassets/blocks/codex/AGENTS.md.blockassets/blocks/claude-code/CLAUDE.md.blockdist/watch/ssrf.jsdist/plugins/loader.jsAGENTS.md~/.codex/config.toml~/.claude/CLAUDE.md~/.claude/skills/*~/.claude/commands/wigolo.md.cursor/rules/wigolo.mdc.github/copilot-instructions.md
Network endpoints10
github.com/searxng/searxng/archive/refs/heads/master.tar.gzduckduckgo.comapi.semanticscholar.orgapi.stackexchange.comapi.search.brave.comapi.github.comwww.bing.comdeveloper.mozilla.orgexport.arxiv.orglocalhost:11434

Decision evidence

public snapshot
AI called this Suspicious at 83.0% confidence as Benign with medium false-positive risk.
Evidence for warning
  • Explicit `wigolo setup mcp`/`wigolo init` can write MCP entries for multiple AI tools in `dist/cli/tui/config-writer.js`.
  • Agent handlers install instruction blocks/skills/commands into AI-agent surfaces, including Codex `AGENTS.md`, Claude skills, Cursor rules, and VS Code instructions.
  • Installed blocks instruct agents to prefer wigolo for web operations, e.g. `assets/blocks/codex/AGENTS.md.block` and `assets/blocks/claude-code/CLAUDE.md.block`.
Evidence against
  • `package.json` has no preinstall/install/postinstall lifecycle scripts.
  • Entrypoint `dist/index.js` only dispatches CLI commands; setup/init are user-invoked, not install-time or import-time mutation.
  • MCP config entries point back to `npx -y wigolo`, a package-aligned local MCP server, not a foreign payload.
  • Network and shell use appears package-aligned: web search/fetch, local SearXNG/Docker/bootstrap, doctor probes, and LLM providers.
  • SSRF guard in `dist/watch/ssrf.js` blocks localhost, private IP ranges, and metadata hostnames for watched URLs.
  • Dynamic plugin loading in `dist/plugins/loader.js` is from configured local plugin directories, not remote code download.
Behavioral surface
Source
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNativeBindingsNetworkShell
Supply chain
HighEntropyStringsUrlStrings
Manifest
CopyleftLicense
scanned 370 file(s), 1.34 MB of source, external domains: 127.0.0.1, api.github.com, api.search.brave.com, api.semanticscholar.org, api.stackexchange.com, api2.marginalia-search.com, devdocs.io, developer.mozilla.org, duckduckgo.com, example.com, export.arxiv.org, github.com, hn.algolia.com, lite.duckduckgo.com, lobste.rs, news.ycombinator.com, python.org, twitter.com, www.bing.com, www.example.com, www.google.com, www.mojeek.com

Source & flagged code

7 flagged · loading source
dist/searxng/docker.jsView file
1import { execSync } from "node:child_process"; L2: import { getConfig } from "../config.js";
High
Child Process

Package source references child process execution.

dist/searxng/docker.jsView on unpkg · L1
dist/plugins/loader.jsView file
84const fileUrl = pathToFileURL(entryPath).href; L85: mod = await import(fileUrl); L86: } catch (err) {
Medium
Dynamic Require

Package source references dynamic require/import behavior.

dist/plugins/loader.jsView on unpkg · L84
dist/search/evidence.jsView file
1import { createHash } from "node:crypto"; L2: import { extractHighlights } from "./highlights.js";
Low
Weak Crypto

Package source references weak cryptographic algorithms.

dist/search/evidence.jsView on unpkg · L1
dist/watch/ssrf.jsView file
1const ALLOWED_PROTOCOLS = /* @__PURE__ */ new Set(["http:", "https:"]); L2: const PRIVATE_HOSTNAMES = /* @__PURE__ */ new Set([ L3: "localhost", ... L69: reason: `${fieldLabel} is not a valid URL`, L70: hint: 'Pass a fully qualified http(s) URL (e.g. "https://example.com/path").' L71: };
High
Cloud Metadata Access

Source reaches cloud instance metadata or link-local credential endpoints.

dist/watch/ssrf.jsView on unpkg · L1
dist/extraction/site-extractors/amazon.jsView file
249contains invisible/control Unicode U+200D (zero width joiner) const cleaned = raw.replace(/[<U+200D><U+200E><U+200F><U+200B>]/g, "").trim();
Critical
Trojan Source Unicode

Source contains bidi control or invisible Unicode characters associated with Trojan Source attacks.

dist/extraction/site-extractors/amazon.jsView on unpkg · L249
Trigger-reachable chain: manifest.main -> dist/index.js -> dist/cli/mcp.js -> dist/server.js -> dist/extraction/pipeline.js -> dist/extraction/v1/site-extractors.js -> dist/extraction/site-extractors/amazon.js Reachable file contains a blocking source-risk pattern.
Critical
Trigger Reachable Dangerous Capability

A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.

dist/extraction/site-extractors/amazon.jsView on unpkg
dist/cli/doctor.jsView file
62try { L63: const r = spawnSync("npx", ["playwright", "--version"], { encoding: "utf-8", timeout: 5e3 }); L64: if (r.status === 0) {
High
Runtime Package Install

Package source invokes a package manager install command at runtime.

dist/cli/doctor.jsView on unpkg · L62

Findings

2 Critical4 High4 Medium6 Low
CriticalTrojan Source Unicodedist/extraction/site-extractors/amazon.js
CriticalTrigger Reachable Dangerous Capabilitydist/extraction/site-extractors/amazon.js
HighChild Processdist/searxng/docker.js
HighShell
HighCloud Metadata Accessdist/watch/ssrf.js
HighRuntime Package Installdist/cli/doctor.js
MediumDynamic Requiredist/plugins/loader.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowWeak Cryptodist/search/evidence.js
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
LowCopyleft License