Integration proposal: Agentegrity — agent integrity and security verification for ADK

[requie]

Agentegrity is an open agent-integrity and security-verification framework from Cogensec with native Google ADK instrumentation.

It adds:

• Agent integrity scoring
• Adversarial-input and prompt-injection detection
• Behavioral drift monitoring
• Reasoning and memory-integrity checks
• Decision provenance
• Cryptographically signed attestations
• Tamper-evident audit chains
• Recovery verification

Agentegrity instruments ADK agents without replacing the ADK runtime or changing the agent architecture.

Repository: https://github.com/Cogensec/agentegrity

Proposed catalog category: Observability / Security

Proposed catalog description:

Verify ADK agent integrity with adversarial detection, behavioral drift monitoring, signed attestations, decision provenance, and security evidence.

The integration is already implemented in the published Agentegrity Python package through its Google ADK adapter.

Installation:

pip install "agentegrity[google-adk]"

Basic usage:

import agentegrity

runtime = agentegrity.init()

secured_agent = runtime.instrument(agent)

# Run the ADK agent normally.

report = runtime.report()

agentegrity.shutdown()