chainlink-confidential-ai-attester-skill

Submits private documents to LLM inference inside an AWS Nitro Enclave and returns attested results.

1|Updated Sep 4, 2025

Search Tags:#kyc #attestation #tee #chainlink #llm inference #confidential ai #defi lending

One-click install

npx skills add https://github.com/FuzzysTodd/The-Nexus-Protocol-Token-DAO --skill chainlink-confidential-ai-attester-skill-fuzzystodd

Or copy as Structured Prompt for Agent▼

Please help me install this Agent Skill.
Skill: chainlink-confidential-ai-attester-skill
Source: https://github.com/FuzzysTodd/The-Nexus-Protocol-Token-DAO/tree/main/skills/chainlink-confidential-ai-attester-skill
Command: npx skills add https://github.com/FuzzysTodd/The-Nexus-Protocol-Token-DAO --skill chainlink-confidential-ai-attester-skill-fuzzystodd

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Applications that need AI analysis of sensitive documents (bank statements, ID documents, brokerage records) cannot send raw data to a public LLM API without exposing user privacy, and on-chain consumers cannot trust an off-chain AI result without cryptographic proof. This Skill runs LLM inference inside a Trusted Execution Environment so raw documents never leave the enclave and results carry a cryptographic attestation. ## Core Features & Use Cases - Confidential inference API: Submit documents via `POST /v1/inference` with base64-encoded resources, then poll `GET /v1/inference/{id}` for the attested result. - Model selection: Choose `gemma4` for images/general tasks or `qwen3.6` for long text, with optional CRE callback webhooks instead of polling. - Prompt templates: Pre-built JSON-enforcing prompts for undercollateralized DeFi lending decisions, accredited investor verification (SEC Rule 501), KYC/AML screening, and proof-of-reserves checks. - Use Case: A lending dApp uploads a user's bank statement, receives an attested approved/denied JSON decision, and verifies on-chain that a specific model ran on specific data — without the financials ever being exposed. ## Quick Start Submit a base64-encoded bank statement to the confidential inference endpoint with a lending-decision prompt, then poll the returned job ID until the attested JSON decision is ready.

Frequently Asked Questions about chainlink-confidential-ai-attester-skill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I run private AI inference on sensitive documents?▼

Submit the document as a base64-encoded resource to POST /v1/inference with your model choice and prompt. The document is processed inside an AWS Nitro Enclave TEE, and you receive an attested result without the raw data leaving the enclave.

How to get an attested AI decision for undercollateralized DeFi lending?▼

Upload a bank statement with a prompt that asks a binary lending question and enforces an exact JSON schema for the approved/denied response. Poll the returned job ID until status is completed, then use the attested output on-chain.

Which model should I use, gemma4 or qwen3.6?▼

Use gemma4 for images and general document analysis; it is the default. Use qwen3.6 when analyzing long text documents that exceed typical context needs.

Can I use a webhook instead of polling for inference results?▼

Yes, include a cre_callback object with your URL in the request body and the result is delivered to that endpoint. If you omit it, poll GET /v1/inference/{id} every 2-5 seconds until the status is completed or failed.

Why is my PDF inference taking so long?▼

PDF preprocessing inside the enclave can take up to 5 minutes. For demos and faster iteration, convert documents to PNG images before submitting them as resources.