Skip to content

Security: StrategicPromptArchitect-AI/MalPromptSentinel-CC-Skill

Security

SECURITY.md

Security Policy

Reporting Security Vulnerabilities

MalPromptSentinel (MPS) is a security tool. We take security issues seriously.


How to Report

Email: StrategicPromptArchitect@gmail.com

Subject Line: SECURITY - MPS Vulnerability Report

Include:

  1. Description of the vulnerability
  2. Steps to reproduce
  3. Potential impact
  4. Suggested fix (if any)

What to Report

Report These Issues:

  • Pattern bypasses - Attacks that evade detection
  • False negatives - Dangerous content marked as safe
  • Code vulnerabilities - Bugs in scanner logic
  • Regex exploits - ReDoS or other regex attacks
  • Information disclosure - Unintended data exposure

Do NOT Report:

  • Known limitations - Evasion techniques (documented)
  • Performance issues - Slow scans, high memory
  • Feature requests - Use contribution process instead

Response Timeline

Action Timeframe
Acknowledgment 48 hours
Initial Assessment 1 week
Fix Development 2-4 weeks
Public Disclosure After fix released

Disclosure Policy

  1. Report privately via email
  2. Allow time for fix development
  3. Coordinate disclosure timing with maintainer
  4. Credit given to reporter (unless anonymity requested)

Known Security Limitations

MPS has documented limitations that are NOT vulnerabilities:

Evasion Techniques (6% detection)

  • Base64 encoding bypasses pattern matching
  • Hex/URL encoding not fully decoded
  • Leetspeak substitution partially detected
  • Unicode obfuscation limited detection

This is by design - Pattern-based detection has inherent limits.

Single-Request Analysis

  • No conversation state tracking
  • Multi-turn attacks not detected
  • Session manipulation limited detection

Mitigation: Use MPS as part of defense-in-depth strategy.


Security Best Practices

When using MPS:

  1. Don't rely solely on MPS - Use multiple security layers
  2. Review ORANGE results - Require human judgment
  3. Always block RED - High confidence attacks
  4. Monitor for bypasses - Report new attack patterns
  5. Keep updated - Use latest version

Scope

This security policy covers:

  • quick_scan.py
  • deep_scan.py
  • mps_patterns.py
  • Supporting scripts

Does NOT cover:

  • Test framework (mps-test-agent)
  • Documentation
  • Example code

Contact

Security Issues: StrategicPromptArchitect@gmail.com

General Questions: Same email, different subject line

Website: https://StrategicPromptArchitect.ca


Recognition

Security researchers who report valid vulnerabilities will be:

  • Credited in CHANGELOG (unless anonymity requested)
  • Acknowledged on website (with permission)
  • Thanked publicly (with permission)

Thank you for helping keep MPS secure!

There aren’t any published security advisories