Security scanner for AI-generated code, right inside your editor. 210 rules tuned for what Cursor, Lovable, Bolt, Replit, and Claude Code actually produce — unprotected webhooks, hardcoded API keys, SQL injection via template literals, missing auth middleware, CORS misconfigurations, and more.
Runs locally. No account. No network. Nothing uploaded.
Features
- Scan on save — every time you save a file, XploitScan checks it for vulnerabilities and surfaces findings in the Problems panel
- Status bar indicator — at-a-glance count of issues in the current file, color-coded by severity
- Rule documentation — each finding links to the rule's full explanation at xploitscan.com/rules
- Configurable noise floor — filter out low-severity findings if you only want to see critical and high issues
- Works offline — the scanner is bundled with the extension, no API calls
Getting started
- Install the extension
- Open any JavaScript, TypeScript, or Python file
- Save — XploitScan runs automatically
- Check the Problems panel (Cmd+Shift+M / Ctrl+Shift+M) for findings
You can also run scans on demand:
XploitScan: Scan Current File— rescan whatever's in the editorXploitScan: Scan Workspace— scan every open document (for full-repo scans, use the CLI:npx xploitscan scan .)XploitScan: Clear Diagnostics— dismiss all findings
All commands are in the command palette (Cmd+Shift+P / Ctrl+Shift+P).
Settings
| Setting | Default | Purpose |
|---|---|---|
xploitscan.scanOnSave |
true |
Auto-scan when a file is saved |
xploitscan.scanOnOpen |
false |
Auto-scan when a file is opened |
xploitscan.severityThreshold |
low |
Hide findings below this severity |
xploitscan.showStatusBar |
true |
Show the XploitScan indicator in the status bar |
What gets flagged
A few examples of what XploitScan catches that most SAST tools miss:
- Stripe / Clerk / Supabase webhook handlers without signature verification
- Hardcoded Anthropic / GitHub / AWS / Supabase service-role keys
- SQL injection via template literals (
db.query(`SELECT ... ${x}`)) — the default pattern AI tools produce dangerouslySetInnerHTMLfed URL query params or localStoragechild_process.exec/spawn({ shell: true })with user input- Next.js API routes without auth middleware or body validation
- CORS wildcards with credentials
- Open redirects via
res.redirect(req.query.x)
All 210 rules are documented at xploitscan.com/rules with OWASP and CWE mappings.
Privacy
Everything runs locally inside VS Code's extension host. Source code never leaves your machine. No telemetry. No account required. Full privacy policy.
Related tools
- CLI —
npx xploitscan scan .for full-repo scans from the terminal - GitHub Action — marketplace listing for CI
- MCP server —
xploitscan-mcpon npm for Claude Desktop / Cursor agents - Web scanner — xploitscan.com/scan for upload-and-scan flow
Same 210-rule engine across all surfaces.
License
MIT.