How does this skill prevent API key leaks?

It runs a specialized script before every file write operation to detect known API patterns and high-entropy strings, blocking the operation if a potential secret is discovered.

How does it stop prompt injection?

It uses the 'Spotlighting' technique to wrap untrusted input in boundaries and scans for instruction-override patterns like 'ignore previous instructions'.

Does it support privacy regulations like GDPR?

Yes, it includes PII detection and masking for sensitive identifiers like emails, SSNs, and credit cards, helping you maintain data privacy standards.

Where are the security logs stored?

Audit logs are saved locally in a structured JSONL format within the .claude/security/audit-logs/ directory for easy querying and reporting.

Does this skill require external dependencies?

No, all core validation scripts are written in Python and rely only on the standard library to ensure they can run in restricted environments.

Security Validation

Name: Security Validation
Author: vanman2024

byvanman2024

0•

Security & Testing

Protects AI workflows by implementing runtime secret scanning, PII masking, and prompt injection defense mechanisms.

The Security Validation skill provides a robust multi-layered defense framework for AI agents, ensuring every interaction remains safe and compliant. It automates critical security tasks such as identifying high-entropy secrets before they are written to disk, masking sensitive PII in user inputs, and detecting sophisticated prompt injection patterns using techniques like Microsoft's Spotlighting. By integrating structured audit logging and data exfiltration prevention, this skill enables developers to deploy AI agents in sensitive environments with confidence, maintaining a clear trail of all security-relevant actions.

Key Features

01Multi-pattern prompt injection defense to prevent instruction overrides

020 GitHub stars

03Structured JSONL audit logging for compliance and forensic analysis

04Automated PII detection and masking for emails, SSNs, and credit cards

05Real-time secret scanning with Shannon entropy analysis to block API key leaks

06Output validation and URL allowlisting to stop data exfiltration attempts

Use Cases

01Preventing accidental commitment of cloud credentials or API keys during code generation

02Hardening AI agent interfaces against adversarial prompt engineering and malicious injections

03Sanitizing sensitive customer data from logs and inputs to maintain GDPR/CCPA compliance

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add vanman2024/dev-lifecycle-marketplace security-validation

For use in Claude.ai and ChatGPT

Download Skill