Arxi Decision Gate Asset Core
Docs

Decision Gate Examples

Decision Gate takes logic, evidence, and gates and turns them into verifiable workflows. Every workflow here is evidence-backed, fail-closed, and auditable.

Agent Task Verification

What This Solves

When an agent reports “done,” you need more than a claim—you need proof. This example shows how Decision Gate turns agent output into a verifiable result you can accept without manual review.

Workflow

  1. Agent performs the task and produces evidence
  2. Decision Gate evaluates the evidence against a gate
  3. If the gate passes, the task is accepted and audited

Spec Snippet (Gate Requirement)

{
  "gate_id": "tests-passed",
  "requirement": { "Condition": "tests_ok" }
}

Gate passes when test failures = 0.

Key takeaway

This pattern turns “done” into “done with proof.” It applies anywhere an agent’s output can be checked against real data—tests, reports, transformations, or compliance checks. The outcome is no longer a claim; it’s verifiable. That verification can drive the loop itself—iterate until the gate passes, or stop when a violation appears.

Release Quality Gate

What This Solves

Before you ship a change, you need proof it meets quality and safety requirements. This example shows how Decision Gate blocks a release unless every required check has real evidence.

Workflow

  1. Your release process runs tests, coverage, and security checks
  2. Each tool writes a JSON result file
  3. Decision Gate evaluates all conditions before release

Spec Snippet (AND Logic)

{
  "gate_id": "quality-checks",
  "requirement": {
    "And": [
      { "Condition": "tests_ok" },
      { "Condition": "coverage_ok" },
      { "Condition": "scan_ok" }
    ]
  }
}

Gate passes when tests failed = 0 and coverage ≥ 85% and critical vulnerabilities = 0.

Key takeaway

This is a release gate that turns many signals into one decision. The same pattern fits any process where multiple checks must agree before action—quality, safety, compliance, or risk. Whether a release is triggered by a human, a pipeline, or an agent, the gate enforces the same rule: you ship only when the evidence says you can.

Human Approval (Quorum)

What This Solves

Some decisions require human sign-off, even when agents do most of the work. This example shows how to require two approvals out of three, without waiting on everyone.

Workflow

  1. An agent prepares the decision and requests approvals
  2. Reviewers submit approvals as evidence
  3. Decision Gate evaluates quorum logic
  4. Once 2 approvals exist, the gate passes and the decision is audited

Spec Snippet (RequireGroup)

{
  "gate_id": "review-quorum",
  "requirement": {
    "RequireGroup": {
      "min": 2,
      "reqs": [
        { "Condition": "alice_approved" },
        { "Condition": "bob_approved" },
        { "Condition": "carol_approved" }
      ]
    }
  }
}

Gate passes when at least 2 of 3 reviewers have approved.

Key takeaway

Quorum logic turns human sign‑off into a rule the system can enforce. It generalizes to any approval flow—legal, security, finance—where you need clear thresholds and an audit trail. The process becomes reliable without requiring unanimous agreement.

Learn More

That completes the introduction to Decision Gate. Compare editions in Decision Gate Features, or go straight to the docs for specs, math, and guarantees.