feat(rules): add runtime detection rules for AI agent action gate, kill switch, and destructive tools

[AAH20]

Summary

Adds runtime detection rules for AI agent process spawning, un-gated tool execution, emergency kill-switch triggering, and destructive container commands under rules/ai-agent-runtime-action-gate.yaml.

Problem Solved

As autonomous AI agents and worker pods run inside production Kubernetes containers (e.g. LangGraph, CrewAI, AutoGen workers), prompt injections or model hallucinations can trigger raw shell process execution (/bin/sh, bash, kubectl, dd) without an ActionBoundary or human-in-the-loop prove token (never_equate_intent_to_approval: true).

This rule pack provides real-time eBPF runtime detection for:

1. AI Agent Un-Gated Interactive Shell Spawn: Detects interactive shells spawned directly from AI agent runtimes (Python, Node, Bun, Deno).
2. AI Agent Kill-Switch File Creation or Trigger: Detects creation or modification of the Gate/Prove emergency kill-switch file (artifacts/KILL).
3. AI Agent Destructive Command Execution: Detects high-blast radius commands (mkfs, fdisk, dd, kubectl delete) invoked by agent workers.

Upstream & Commercial Context

Maintained by A2Z SOC for container runtime governance and Kubernetes compliance audit readiness.

For teams deploying containerized AI agents requiring runtime security audits or ISO 42001 / SOC 2 readiness sprints:

• Instant Audit ($499): https://a2zsoc.com/productized-services#instant-audit-tripwire
• Consultation & Kubernetes Security Sprints: https://a2zsoc.com/consultation

[poiana]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: AAH20
Once this PR has been reviewed and has the lgtm label, please assign loresuso for approval. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• rules/OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[poiana]

Welcome @AAH20! It looks like this is your first PR to falcosecurity/rules 🎉