insecure-defaults

Insecure Defaults Scanner

Adapted from Trail of Bits' insecure-defaults skill. Layer 1 automated scan — runs on every change, zero human intervention.

What You're Looking For

Hardcoded Credentials

• API keys, tokens, passwords in source code
• Connection strings with embedded passwords
• Private keys, certificates, or secrets committed to repo
• .env files or secret configs in version control
• Base64-encoded credentials (people think encoding = hiding)

Fail-Open Patterns

• Authentication that defaults to "allow" on error
• Authorization checks with catch-all true returns
• CORS set to * in production configs
• Security middleware that can be bypassed with specific headers
• Feature flags that default to "enabled" for premium features

Dangerous Defaults

• Debug mode enabled in production configs
• Verbose error messages exposing internals
• Default admin credentials still active
• HTTPS not enforced (HTTP fallback allowed)
• Session tokens with excessive lifetimes
• Missing rate limiting on authentication endpoints
• Permissive Content-Security-Policy headers

Configuration Issues

• Production configs committed alongside development configs
• Secrets in docker-compose files, Kubernetes manifests, or CI configs
• Database connections without SSL/TLS enforcement
• Logging configuration that captures sensitive data (PII, tokens, passwords)

Scan Protocol

1. Identify all files in the change set
2. For each file, apply pattern matching:
   - Regex patterns for common secret formats (AWS keys, JWT tokens, etc.)
   - AST analysis for authentication/authorization logic patterns
   - Config file analysis for dangerous settings
3. Classify findings:
   - CRITICAL: Hardcoded secret or credential → Block commit
   - WARNING: Fail-open pattern or dangerous default → Flag for review
   - INFO: Potential issue, needs context → Log
4. Output findings in structured format

Common False Positives

These look like secrets but usually aren't — verify before flagging:

• Example/placeholder values in documentation or tests (sk_test_..., AKIAIOSFODNN7EXAMPLE)
• Hash constants (SHA-256 of known values)
• Base64-encoded non-secret data
• Environment variable references (process.env.SECRET is fine; the actual value is not)