Works with the AI tools you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIGitHub CopilotGitHub CopilotGemini CLIGemini CLI+17 more

    ZeroDay Scanner

    2

    Read-only Windows scan for supply-chain worms and AI-assistant persistence that antivirus misses — with wiper-safe cleanup.

    Secure checkout via Stripe

    0 installsSecurity scanned

    See it in action

    You say

    Scan my Windows machine for supply-chain worms and hijacked AI-assistant configs, then show me anything suspicious: powershell -ExecutionPolicy Bypass -File .\scripts\scan.ps1

    Your agent does

    Result: 2 FLAG(S) | Check | Status | Detail | | py-startup | PASS | No sitecustomize/usercustomize startup hooks | | compromised-pkgs | FLAG | PyPI 'rlask' found in requirements.txt (known typosquat) | | ai-config | PASS | Claude/Cursor/VS Code hooks unchanged, no exfil patterns | | mcp | PASS | 52 MCP server defs across 14 configs, all known/clean | | skill-integrity | FLAG | Hidden Unicode (prompt-injection vector) in skill 'foo' | | persistence | PASS | Run keys, Startup, tasks, services match baseline |

    Report: reports\scan-20260710-1421.md (exit code 1)

    What you get

    Detect malicious .pth files and Python startup hooks in site-packages.Audit AI assistant configs for rogue tasks and MCP server commands.Identify hidden prompt-injection signals in local skills and plugins.Execute a safe quarantine-first cleanup to prevent malware retaliation.Scan a dev machine after installing a suspicious npm or pip packageWeekly scheduled supply-chain / persistence check on a workstationAudit an AI-dev machine for hijacked Claude/Cursor/VS Code hooks and rogue MCP serversPost-incident triage: find droppers, .pth hooks, and exfil GitHub reposVerify a laptop is clean before rotating credentialsDetect tampered agent skills/plugins and invisible-Unicode prompt injection

    About this skill

    ZeroDay is a read-only Windows scanner that hunts the new wave of supply-chain and AI-assistant attacks traditional antivirus misses: self-propagating npm/PyPI worms (Shai-Hulud, Miasma/TeamPCP class), malicious Python startup hooks (.pth / sitecustomize), hijacked AI coding-assistant configs (Claude, Cursor, VS Code — hooks, tasks.json, MCP servers), tampered agent skills/plugins, prompt-injection via invisible Unicode, credential-exfil GitHub repos, and stealth persistence (Run keys, Startup, scheduled tasks, services).

    It runs 11 checks against a baseline you capture once, writes a clear markdown report plus a live self-refreshing status dashboard, and never modifies your system while scanning. When you confirm a threat, a separate, wiper-safe cleanup tool (quarantine.ps1) preserves a forensic copy first, then optionally neutralizes it (reversible) or removes it — and a plain-English REMOVAL.md walks you through the correct isolate-first order that avoids triggering a worm's file-wiper retaliation.

    ZeroDay improves itself over time. Supply-chain campaigns churn constantly, so its detection signatures live in an updatable feed (signatures.json), not buried in code. A built-in research-and-merge workflow finds threats disclosed since your last update, verifies each indicator against authoritative advisories (GitHub Advisory Database, OSV / OpenSSF malicious-packages, npm/PyPI, Socket, Microsoft, StepSecurity), and merges only high-confidence, low-false-positive indicators — new worm packages, dropper filenames, exfil hosts, and repo-name patterns — via update-signatures.ps1, which validates and de-duplicates every entry so a bad signature can never flag everything. The scanner unions these with built-in fallbacks and reminds you when signatures go stale, so coverage keeps growing without ever touching the tool's logic.

    No install, no dependencies beyond PowerShell, no admin required, and no data leaves your machine. Every release is sealed with a SHA-256 integrity manifest (verify-integrity.ps1) so you can confirm the download is untampered before you trust it.

    How to install

    Drop the file into your AI Agent. Works with Claude, Cursor, ChatGPT, and 20+ more.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    30-day refund

    Not a fit? Get your money back

    Trust & safety

    Security scanned

    Verified clean 1 month ago

    • One-time purchase, yours forever

    Listed1 month ago

    Frequently Asked Questions