🔍 sherscan

AI-aware security scanner, finds exposed secrets in your codebase and checks whether your AI coding agent (Cursor, Copilot, Windsurf, Claude Code, etc.) can actually read your sensitive files.

Most secret scanners stop at "you have a .env file." sherscan goes one step further: it checks whether that file is actually blocked from AI agents via .cursorignore, .aiexclude, .clineignore, Claude Code's deny rules.

Why

AI coding assistants can read almost any file in your project by default. If your .env, private keys, or credentials aren't explicitly excluded, your AI agent (and anything it shares context with) can see them. sherscan audits this blind spot in seconds.

Features

• 🔎 Sensitive file detection, flags .env files, private keys (.pem, .key, id_rsa), cloud credentials (.aws/credentials, .npmrc), keystores, and more.
• 🔑 Secret scanning, greps file contents for real leaked credentials: AWS keys, GitHub tokens, Stripe keys, Supabase keys, Anthropic keys, SendGrid keys, private key blocks, JWTs, and generic hardcoded secrets.
• 🤖 AI exposure check, cross-references every sensitive file against the ignore files of 9+ AI tools (Cursor, Windsurf, Cline, Gemini, Codeium, Copilot, Claude Code, and generic .aiignore) to see if it's actually protected.
• 🛡️ Masked previews, secrets are shown masked (e.g. AKIA********MNOP) so reports are safe to share/paste.
• 📊 Readable or machine output, pretty colored terminal report by default, or --json for scripting.
• 🚦 CI-friendly, --ci flag exits with code 1 when exposed secrets are found, so you can fail a pipeline on real risk.

Installation

Run it directly without installing anything (recommended):

npx sherscan

Or install globally:

npm install -g sherscan

Or add it as a dev dependency to run via npm scripts:

npm install --save-dev sherscan

Usage

Scan the current directory:

npx sherscan

Scan a specific project path:

npx sherscan ./path/to/project

Output JSON (for scripting or piping into other tools):

npx sherscan --json

Fail CI when a sensitive file with no AI-ignore coverage is found:

npx sherscan --ci

Options

| Flag | Description | | --------------- | ------------------------------------------------------------- | | [path] | Project directory to scan (defaults to .) | | --json | Output results as JSON instead of a colored report | | --ci | Exit with code 1 if any fully-exposed findings are detected | | -V, --version | Print the version number | | -h, --help | Print usage help |

Example output

────────────────────────────────────────────────────
  🔍 sherscan  · AI-aware security scanner
────────────────────────────────────────────────────

────────────────────────────────── SUMMARY ──────────────────────────────────

  Sensitive files found       1
  Secrets in content          1
  Files exposed to AI agents  1

    HIGH      1 file(s)

────────────────────────────── SECRETS DETECTED ─────────────────────────────

    HIGH    .env
           Line 1  ·  AWS Access Key
           Preview: SECRET_KEY=AKIA********MNOP

──────────────────────────────── AI EXPOSURE ────────────────────────────────

    HIGH    .env
           Status   fully exposed   has secret

──────────────────────────────── RECOMMENDATIONS ────────────────────────────

  ✖  No AI ignore files found in this project

  →  Create a .aiignore or .cursorignore and add:
       .env

  ⚠   Rotate these credentials immediately:
      .env

────────────────────────────────────────────────────
  ✖  Issues found — review recommendations above
────────────────────────────────────────────────────

What it checks

Sensitive files

sherscan flags files matching common sensitive-file patterns, including:

• .env, .env.* (any environment file)
• *.pem, *.key, id_rsa*, id_ed25519* (private keys)
• *.p12, *.pfx, *.keystore (certificates/keystores)
• credentials.json, secrets.*
• .npmrc
• .aws/credentials, .aws/config
• Supabase's supabase/.temp/** directory

Secret patterns

File contents (for common source/config file types under 500 KB, skipping binaries) are scanned for:

• AWS Access Keys
• GitHub Tokens
• Stripe Live/Test Keys
• Supabase Keys
• Anthropic API Keys
• SendGrid Keys
• Private key blocks (-----BEGIN ... PRIVATE KEY-----)
• Generic JWTs
• Hardcoded SECRET / API_KEY / PRIVATE_KEY / AUTH_TOKEN assignments

Matches inside comments are skipped, and all detected secrets are masked in the report.

AI exposure

For every sensitive file found, sherscan checks whether it's covered by any of these AI-tool ignore mechanisms:

| Tool | Ignore file | | ----------- | -------------------------------------------------- | | Cursor | .cursorignore, .cursorindexingignore | | Windsurf | .windsurfignore | | Cline | .clineignore | | Gemini | .aiexclude, .geminiignore | | Codeium | .codeiumignore | | Copilot | .copilotignore | | Claude Code | .claude/settings.json (permissions.deny rules) | | Generic | .aiignore |

A file is marked fully exposed if none of the detected AI tools' ignore rules cover it — meaning an AI agent working in that project could read it.

Using in CI

Add sherscan to your pipeline to catch exposed secrets before they ship:

# .github/workflows/security.yml
name: Security check
on: [push, pull_request]
jobs:
  sherscan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npx sherscan --ci

The command exits with code 1 if any sensitive file is fully exposed to an AI agent, failing the build.

Requirements

• Node.js >= 18

License

MIT