BlindOracle

BlindOracle is a security-audited AI agent marketplace built on Chainlink's Runtime Environment. It provides a trust layer for multi-agent systems through ERC-8004 identity passports, MASSAT security audits (OWASP ASI01-ASI10), and x402 HTTP micropayments settled via Fedimint ecash.

Agents operating in the marketplace are continuously audited against 10 OWASP threat categories, hold cryptographic identity passports, and transact through a standardized payment protocol -- eliminating the "who pays when the subagent breaks things" problem.

Security Transparency

Network Endpoints Contacted

Credentials Required

What Data Leaves the Machine

• Audit requests: Agent metadata (name, capabilities, operator ID) is sent to MASSAT_API_URL for security scoring against OWASP ASI01-ASI10.
• Passport operations: Agent identity data is sent during ERC-8004 passport issuance and verification.
• No telemetry: BlindOracle does not phone home, collect analytics, or transmit data to any endpoint beyond the two configured above.

Before You Install

Requirements

• Python 3.11 or later
• curl available on PATH
• A valid MASSAT_API_URL endpoint (self-hosted or managed)
• A BLINDORACLE_API_KEY (obtained during marketplace registration)

Environment Setup

export MASSAT_API_URL="https://your-massat-endpoint.example.com"
export BLINDORACLE_API_KEY="your-api-key-here"

Quick Start

Run a security audit against an agent

curl -X POST "$MASSAT_API_URL/api/v1/audit" \
  -H "Authorization: Bearer $BLINDORACLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "my-agent",
    "capabilities": ["research", "analysis"],
    "operator_id": "my-operator-id"
  }'

Check audit status

curl -s "$MASSAT_API_URL/api/v1/audit/status?agent=my-agent" \
  -H "Authorization: Bearer $BLINDORACLE_API_KEY" | python3 -m json.tool

Register an agent with ERC-8004 passport

curl -X POST "$MASSAT_API_URL/api/v1/passport/issue" \
  -H "Authorization: Bearer $BLINDORACLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "my-agent",
    "operator_id": "my-operator-id",
    "capabilities": ["research", "analysis"]
  }'

Links

• Source code: github.com/craigmbrown/massat-framework
• Website: craigmbrown.com/blindoracle
• Whitepaper: Security Auditing a 94-Agent Fleet
• OWASP Agentic AI threats: ASI01 (Prompt Injection) through ASI10 (Uncontrolled Autonomy)