Tutorials, Demos & Guides
Docs » Tutorials, Demos & Guides » ZQK Core Interactive Demonstrations & Visual Showcase
Raw .md

ZQK Core Interactive Demonstrations & Visual Showcase

Principle: Agents manage the work. ZQK enforces the physics.

The demonstrations in scripts/demos/ showcase the core architectural guarantees of the Zen Quantum Kernel (ZQK). Unlike conventional AI coding tools where agent memory is volatile, execution state is lost on crashes, and security is left to LLM prompt compliance, ZQK operates as a durable, cellular operating system and knowledge kernel.

Every demo in this suite performs 100% authentic operations against the live kernel, with zero simulated mocks.

Quickstart: Running the Demonstrations

# Run the interactive menu (select 1-5, or run all)
./scripts/demos/run-all-demos.sh

# Run all 5 demos in non-interactive batch mode
./scripts/demos/run-all-demos.sh --batch

# Run all 5 demos and capture fresh high-fidelity SVG screenshots
./scripts/demos/run-all-demos.sh --batch --capture

# Run a specific demo directly
./scripts/demos/day_zero_dx_demo.sh
./scripts/demos/time_travel_demo.sh
./scripts/demos/containment_breach_demo.sh
./scripts/demos/crypto_audit_demo.sh
./scripts/demos/object_inspector_policy_studio_demo.sh

Demo 1: Day-0 Greenfield DX & Instant Agent Orientation

Thesis: Infrastructure should feel effortless. A developer or agent swarm must be able to initialize a brand-new project and achieve full knowledge-kernel governance and autonomous orientation in seconds.

Key Capabilities Demonstrated:

  • Zero-Config Bootstrap: zqk system init --project-name <name> in an isolated workspace.
  • Agent Host Detection & Seating: zqk system agent-onboard detects the IDE (Cursor, VS Code, Windsurf, Cline) and seeds vendor-neutral seating directives.
  • Instant CAS Object Minting: Questions, goals, and tasks are immediately indexed into Content-Addressable Storage (CAS).
  • Autonomous Next-Action Discovery: Agents run zqk workflow whats-next to discover their orientation without asking a human "what should I do?".
  • Zero-Defect System Health: zqk system check validates the CAS membrane, process locks, and storage integrity out-of-the-box in < 3 seconds.

Demo 1: Day-0 Greenfield DX

Demo 2: The "Kill -9" Resurrection & Crash Consistency

Thesis: Most autonomous AI agents keep state in volatile RAM. When killed, context dies. ZQK treats agent state as a durable, transaction-isolated cellular microkernel.

Key Capabilities Demonstrated:

  • Durable CAS Persistence: Graph mutations and seat credentials are written to durable storage before acknowledgment.
  • Unannounced SIGKILL (kill -9): A host process holding an active write lock is forcefully terminated mid-flight with kill -9.
  • Autonomous Self-Remedy: zqk system check --auto-remedy autonomously detects abandoned PIDs and stale locks, cleans them, and validates all SHA-256 CAS object leaves.
  • Zero Data Loss: All state, keystore credentials, and plan items resume with 100% integrity without requiring re-prompting.

Demo 2: Kill -9 Resurrection

Demo 3: Dynamic Membrane & Fail-Closed Prompt Injection Intercept

Thesis: Holon as Boundaries. An LLM cannot be trusted to self-regulate its host machine access. In ZQK, the kernel—not the model—enforces the physics.

Key Capabilities Demonstrated:

  • Adversarial Bait Scenario: An agent is instructed via a hidden prompt injection to exfiltrate host secrets (/etc/passwd, ~/.ssh/id_rsa) or write outside the project root (../../../../tmp/pwned.txt).
  • Live MCP Server Connection: Real JSON-RPC client session connected to zqk mcp serve.

  • Fail-Closed Membrane Intercept: The kernel intercepts the path traversal before filesystem access, returning: ✖ Error: access denied: path escapes sandbox root: path escapes project root

  • Invariant Verification: Automated test suite verifies 11 path-traversal attack vectors fail closed.

Demo 3: Dynamic Membrane Intercept

Demo 4: Verifiable Cryptographic Ledger & CISA OpenVEX Attestation

Thesis: Anyone can generate code. ZQK generates cryptographically verifiable proof of how the code was made. You don't have to trust the LLM; you can mathematically verify the exact constraint and seating path taken.

Key Capabilities Demonstrated:

  • Multi-Agent Seating Issuance: zqk keystore issue assigns cryptographic API credentials, seating tokens, and SHA-256 fingerprints to distinct agent roles (e.g. Swarm-Architect, Swarm-Security-Reviewer).
  • Protected Vault Keystore: Agent credentials and cryptographic hashes are stored in the isolated kernel vault.
  • CISA OpenVEX v0.2.0 Attestation: Automatically generates signed, machine-readable vulnerability and supply-chain attestations.
  • Independent Verification: Cryptographic verification gate confirms OpenVEX schema conformance and tamper evidence.

Demo 4: Cryptographic Ledger & OpenVEX

Demo 5: Interactive Object Inspector, Live Policy Studio & Unified QA

Thesis: Managing complex knowledge kernel graphs should not require memorizing dozens of CLI flags or parsing walls of raw YAML. Humans get clean, interactive drill-downs and policy studios; autonomous AI agents receive dense, token-optimized semantic projections.

Key Capabilities Demonstrated:

  • Dual Human/Agent Projections: Token-optimized JSON (-f json) for LLM context windows, and structured Terminal Design System (TDS) cards for human operators.
  • Modular Cards: CAS storage profiles (hash, size, permissions), ontology traits, and unbroken lineage radar.
  • Live Policy Rule Studio: Real-time DSL dry-run evaluation (--policy-studio) checking governance rules against live objects.
  • Unified Mission Control QA: zqk test dashboard --check-dod verifies Definition of Done (DoD) compliance and test-to-criteria lineage across the entire graph.

Demo 5: Object Inspector & Policy Studio