hostinger-api-mcp@3.26.0
MCP server for Hostinger API
AI Security Review
scanned 1d ago · by lpm-firewall-aiThe new Headless workflow mandates runtime reporting to Hostinger without prior consent or an opt-out, and its reports include identifying workflow details and user feedback. This is suspicious telemetry; the reviewed source does not show credential theft or transmission to a foreign host.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
When an agent follows the Headless build skill, including at early stops and during later retries of unsent reports.
Impact
Hostinger receives account-associated workflow and feedback data without prior consent or an opt-out in the skill.
Mechanism
The skill forbids asking for consent; run-log marks consent as sent and records domain, tool/model/runtime details, timings, issues and feedback for submission through the Hostinger API.
AI rationale
The new Headless workflow mandates runtime reporting to Hostinger without prior consent or an opt-out, and its reports include identifying workflow details and user feedback. This is suspicious telemetry; the reviewed source does not show credential theft or transmission to a foreign host.
Evidence
skills/headless/SKILL.mdskills/headless/run-log.mjsskills/headless/steps/09-done/SKILL.mdsrc/core/tools/all.js
Network endpoints1
developers.hostinger.com/api/hosting/v1/headless/agent-reports
Decision evidence
public snapshotAI called this Suspicious at 96.0% confidence as Unknown with medium false-positive risk.
Evidence for warning
- The Headless skill requires sending every build report to Hostinger and explicitly says not to ask for consent or offer a choice.
- The report is marked as sent with consent before upload.
- Reports include the domain, tool, model, OS, Node version and MCP details.
- Reports include logged user feedback.
- The upload tool submits reports to Hostinger so it can learn where builds get stuck.
Evidence against
- The report upload is part of the Headless build workflow, not an install hook or default MCP startup action.
- The report goes to Hostinger's own API and is associated with the authenticated account; the reviewed code does not show credentials sent to a foreign host.
- The skill says to paraphrase feedback and not include long quotes, credentials or personal data.
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStrings
Source & flagged code
4 flaggedbuild.jsView file
2L3: import { exec } from 'child_process';
L4: import fs from 'fs';
High
18console.log('Compiling TypeScript...');
L19: exec('npx tsc', (error, stdout, stderr) => {
L20: if (error) {
High
Runtime Package Install
Package source invokes a package manager install command at runtime.
build.jsView on unpkg · L18src/core/skills.jsView file
•matchType = previous_version_dangerous_delta
matchedPackage = hostinger-api-mcp@3.24.0
matchedIdentity = npm:aG9zdGluZ2VyLWFwaS1tY3A:3.24.0
similarity = 0.864
summary = stored previous version shares package body but lacks this dangerous source file
Critical
Previous Version Dangerous Delta
This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
Findings
1 Critical3 High4 Medium4 Low
CriticalPrevious Version Dangerous Deltasrc/core/skills.js
HighChild Processbuild.js
HighShell
HighRuntime Package Installbuild.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
MediumAi Review Evidenceskills/headless/SKILL.md
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to hostinger-api-mcp@3.26.0.
See all versions of hostinger-api-mcp for other recorded verdicts.
- Review the evidence and your use of hostinger-api-mcp@3.26.0 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.