Verified Task

Verify any agent output before it is trusted, paid for, or acted on.

---

🧠 The Problem

AI agents generate outputs — but:

• there is no guarantee the output matches the task
• incorrect results can silently pass through
• payments can be triggered without verification
• downstream systems rely on unverified data

Most workflows assume correctness.

That assumption breaks systems.

---

✅ The Solution

Verified Task is a universal verification pattern for agent workflows.

It ensures that:

• outputs are validated against a defined specification
• results are deterministically verified
• proof is generated before any action is taken
• payment and execution only proceed on success

---

🔁 How It Works

1. Define the task specification
2. Generate the output
3. Verify the output using SAR
4. Receive a verdict + receipt
5. Proceed only if PASS

---

🔍 What You Get

• PASS / FAIL / INDETERMINATE verdict
• Signed SAR receipt (verifiable proof)
• Safe-to-proceed signal
• TrustScore update per verified task

---

⚙️ Example

Input

{
  "task_id": "task-001",
  "agent_id": "0xYourWallet:your-agent",
  "task_spec": "Summarize x402 news in 3 bullet points",
  "expected_output_format": "JSON array of 3 strings",
  "output": ["...", "...", "..."],
  "counterparty": "0xOptionalCounterpartyWallet"
}

Verification Call

settlement_witness({
  task_id: "task-001",
  agent_id: "0xYourWallet:your-agent",
  spec: { expected: task_spec },
  output: { result: agent_output }
})

Endpoint: POST https://defaultverifier.com/settlement-witness

Output

{
  "verdict": "PASS",
  "safe_to_proceed": true,
  "sar_receipt": {
    "receipt_v0_1": {
      "verdict": "PASS",
      "confidence": 1.0,
      "reason_code": "SPEC_MATCH",
      "receipt_id": "sha256:...",
      "verifier_kid": "sar-prod-ed25519-02",
      "sig": "base64url:..."
    },
    "trustscore_update": {
      "score": 54,
      "tier": "bronze",
      "receipts_to_next": 10
    }
  }
}

---

🔒 Why This Matters

Without verification:

• agents produce incorrect results undetected
• payments settle on unverified work
• trust is assumed instead of proven

With Verified Task:

• outputs are provable
• workflows become deterministic
• trust becomes explicit and auditable

---

🧩 Use It Anywhere

Verified Task wraps any workflow:

• content generation
• data processing
• API results
• automation pipelines
• agent-to-agent transactions

If an output matters — verify it.

---

🔐 Offline Verification

Every SAR receipt is independently verifiable:

1. Fetch public key registry:
   https://defaultverifier.com/.well-known/sar-keys.json

2. Match verifier_kid to registry key

3. Canonicalize signed core (RFC 8785)

4. SHA256 hash → Ed25519 verify

No callbacks required. Receipt remains valid forever.

TypeScript SDK: https://www.npmjs.com/package/sar-sdk
Protocol docs: https://sarprotocol.org
SAR spec: https://defaultverifier.com/spec/sar-v0.1

---

🌐 Part of the Agent Commerce Stack

x402 → SettlementWitness → TrustScore
Payment → Delivery verification → Reputation

Verified Task is the standard pattern that connects agent workflows to this verification layer.

---

🔗 Relationship to SettlementWitness

Verified Task is a standard usage pattern for SettlementWitness (SAR).

• SettlementWitness performs verification
• Verified Task makes it easy to apply

This repository defines how to use SAR — not replace it.

---

🚀 The Default Pattern

If an agent is going to:

• get paid
• trigger execution
• make a decision

→ wrap it in Verified Task first.

---

📦 Installation

Direct endpoint

POST https://defaultverifier.com/settlement-witness

MCP server

https://defaultverifier.com/mcp

TypeScript SDK

npm install sar-sdk

SkillsMP / Claude Skills Search: Verified Task

---

📄 Resources

• SAR Protocol: https://sarprotocol.org
• SettlementWitness: https://defaultverifier.com
• SAR Spec: https://defaultverifier.com/spec/sar-v0.1
• Key Registry: https://defaultverifier.com/.well-known/sar-keys.json
• GitHub: https://github.com/nutstrut/default-settlement-verifier
• X: @defaultsettle

---

🏷️ Tags

ai-agents · verification · workflow-safety · agent-commerce ·
trust · automation · sar-protocol · sarprotocol ·
settlementwitness · deterministic-verification ·
x402 · ed25519 · agent-safety · output-validation

---

📋 License

MIT

Metadata

Last updated: 2026-04-02