integrity-check

To scan for prompt injection and memory poisoning, run an integrity check on your project state files. This validates persisted state and agent outputs against adversarial threats using glob, grep, and git-blame for security validation.

AI state integrity verification safeguards persistent state and outputs from malicious tampering. You need it before committing changes to automatically scan `.rune/` files and prevent subtle prompt injection attacks.

Detect identity spoofing and adversarial attacks by scanning agent outputs and context bus data. The integrity check validates these outputs against known threat models to ensure AI decision-making processes remain secure.

Yes, you can validate persisted state in `.rune/` files without external dependencies. The integrity check uses built-in scripts to scan files directly, requiring no additional environment setup or external libraries.

Git-blame works for security validation by tracking unexpected modifications in context bus data. It analyzes file history alongside glob and grep scans to identify hidden instructions and adversarial content.

An integrity check fails on unexpected modifications when it detects adversarial threats like memory poisoning or identity spoofing. It validates skill outputs against known threat models to flag hidden instructions or malicious tampering.