Skip to main content
skill CX Codex

chainlink-confidential-ai-attester-skill — Claude Code skill

chainlink-confidential-ai-attester-skill runs LLM inference over a private document inside a TEE and hands back the result; the raw file stays in the enclave. Built for KYC, investor checks and proof-of-reserves.

MIT 226 installs ★ 130 updated 2026-10-23

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-confidential-ai-attester-skill -a claude-code
Open on GitHub

What chainlink-confidential-ai-attester-skill does

The skill wires your agent to the Confidential AI Attester HTTP API. A document goes into the service's Nitro Enclave, the model reads it there, and only text comes back out. Submission is POST /v1/inference with Authorization: Bearer $API_KEY; the key comes from an environment variable or a secret store, never from the command line, and the base URL lives in BASE_URL.

A submit answers 202 Accepted with an id and a queued status. The skill then polls GET /v1/inference/{id} every 2–5 s until completed or failed, and pulls output, usage and completed_at from the finished job. Long text is steered to qwen3.6, images and general work to gemma4. PDFs can take minutes to preprocess, so demos move faster with PNG.

The API exposes no Chainlink signature and no TEE attestation document, so hashing the response fields only shows later edits. It does not prove the inference ran in an enclave. EVM delivery means checking the result against a JSON schema, building a CRE report and sending it through the Chainlink Forwarder; a failed or non-JSON result produces no report. Access is beta: keys come from the Chainlink booth at EthGlobal NYC or the #partner-chainlink Discord channel.

What it helps with

  • run an LLM over a loan file without uploading it publicly
  • pull JSON fields out of a scanned PDF for accredited-investor checks
  • poll an async job by id and collect the output
  • deliver a checked result to an EVM contract via Chainlink Forwarder
  • choose between the attester, a local model and redaction

Example prompts for the agent

  • Submit doc.pdf to POST /v1/inference with gemma4 and keep the id.
  • Poll GET /v1/inference/{id} until it reports completed, then print output.
  • Write the system prompt and a user prompt with an exact JSON schema for the fields I need.

How to install chainlink-confidential-ai-attester-skill

1Skills CLI (Vercel)

npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-confidential-ai-attester-skill -a claude-code

2Manually: put the files in place

~/.claude/skills/chainlink-confidential-ai-attester-skill/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/chainlink-confidential-ai-attester-skill/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about chainlink-confidential-ai-attester-skill

How do I install chainlink-confidential-ai-attester-skill in Claude Code?

Run npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-confidential-ai-attester-skill -a claude-code. The skill file lands at ~/.claude/skills/chainlink-confidential-ai-attester-skill/SKILL.md.

What does chainlink-confidential-ai-attester-skill do?

It sends a document to the Chainlink Confidential AI Attester for inference inside a TEE, polls the job, and returns the resulting text. The source file never leaves the enclave.

Can I prove the inference ran inside a TEE?

Not from the API response. It carries no Chainlink signature and no attestation document; a hash of the fields detects edits afterwards but does not show the work happened in an enclave.

Source text

SKILL.md · MIT
---
name: chainlink-confidential-ai-attester-skill
description: "Runs LLM inference over a private document inside a TEE, returns an attested result, and keeps the raw document in the TEE. No document upload means chainlink-cre-skill. It applies to private inference, attested AI, confidential AI, and sensitive-document analysis, including lending, accredited-investor, KYC/AML, and proof-of-reserves checks."
license: MIT
compatibility: Designed for AI agents that implement https://agentskills.io/specification, including Claude Code, Cursor Composer, and Codex-style workflows.
allowed-tools: Read WebFetch Write Edit
metadata:
  version: "0.0.6"
---

# Chainlink Confidential AI Attester

Private-document LLM inference runs in a service-managed TEE; raw documents never leave it and are never stored or exposed.

Scope: only the Chainlink Confidential AI Attester document-inference HTTP API. Data Feeds, CCIP, AI architecture or zkML, Anthropic or other vendor PDF APIs, and AWS Nitro/Bedrock attestation stay with their owning product or vendor; do not introduce Attester. For any request outside Attester document inference, state that this capability does not apply, route to the owning capability, and do not provide any Attester API or playground procedure.

**EthGlobal NYC hackathon beta.** Get an API key at the **Chainlink booth** or in Discord's **#partner-chainlink channel**. The playground at `https://confidential-ai-dev-preview.cldev.cloud/playground` and the HTTP API are two access paths to the same Attester service, not separate privacy options.

## Workflow 1 — Submit: `POST /v1/inference`

Use `Authorization: Bearer $API_KEY`; load the key from an environment variable or injected secret store, never hardcode it or place its literal or expanded value in command-line arguments. Set the service URL in `BASE_URL` (or an equivalent environment variable) and use it in every command or code sample; never inline the service base.

```json
{
  "model": "gemma4",
  "system_prompt": "",
  "prompt": "...",
  "resources": [{ "filename": "doc.pdf", "content_type": "application/pdf", "content_base64": "<base64>" }],
  "cre_callback": { "url": "https://..." }
}
```

Omit `cre_callback` when polling, including in the playground; use `GET /v1/inference/{id}`. Product guidance must name the playground above and choose `qwen3.6` for long text; `gemma4` is the images/general default. Prefer PNG for demos; PDF preprocessing can take 5 minutes.

Success is `202 Accepted` with `{ "id": "...", "status": "queued" }`; save the `id`.

For any submit or poll answer, emit runnable curl or code—not a completion summary—with the request fields, `202` shape, saved `id`, `GET /v1/inference/{id}`, both terminal statuses, and `output` retrieval (or `error` on failure).

See the [curl example and language guidance](references/code-examples.md) and [API specification](references/api-reference.md).

## Workflow 2 — Poll: `GET /v1/inference/{id}`

Poll every 2–5 s until `completed` or `failed`. Completion fields are `output` (LLM text), `usage`, and `completed_at`; see [troubleshooting](references/troubleshooting.md).

## Verification Boundary and EVM Delivery

The Attester API does not expose an independently verifiable Chainlink signature or TEE attestation document. Hashing response fields can detect a later change to those fields, but a response hash is neither an independently verifiable Chainlink signature nor a TEE attestation and does not prove that inference ran in an enclave.

For EVM delivery, validate a completed result and its required JSON schema, create a CRE-native report, and deliver that report through the Chainlink Forwarder. Treat failed or non-JSON output as failure and create no report. Do not use a generic EIP-712 signer or application private key to present Attester output as confidential-compute proof.

Every signed-summary or verify-later answer must state one credential rule: `API_KEY` comes from an environment variable or injected secret store, never a CLI argument.

## Writing Prompts That Work

Enforce JSON in two layers: (1) the default system prompt unless a change is necessary and (2) a user prompt with a fact-extraction task and exact JSON schema. Emit the requested `system_prompt`, user `prompt`, and schema in full; never replace them with a summary. See the [four use-case templates](references/prompts.md).

## Choose the Privacy Approach

| Option | Use when | Tradeoff |
|--------|----------|----------|
| Attester | You want managed document inference inside the service's Nitro Enclave | Fastest integration, but the current API provides no independently verifiable Chainlink or TEE signature |
| Local model | Data must remain on a controlled device with no external inference call | You operate the model and hardware; locality alone provides no remote attestation |
| Self-hosted confidential compute | You need infrastructure control and can operate a TEE and its attestation verification | Most control and potential for independent attestation, with the highest operational burden |
| Redaction or tokenization | The task can work on minimized or substituted data before any external call | Simplest data minimization, but may reduce utility and does not eliminate re-identification risk |

The playground and API are interfaces to one Attester service; choose between them for interaction style, not as distinct privacy architectures.

© smartcontractkit · MIT · smartcontractkit/chainlink-agent-skills

Similar skills

All skills for AI and ML →
venice-text-routing veniceai venice-text-routing routes prompts to Venice text models based on privacy tiers, modality, context limits, and cost. Helps local agents offload heavy or sensitive prompts to the right cloud model. CX Codex +1 ★ 176repo stars azure-confidential-computing MicrosoftDocs Skill with expert knowledge on Azure Confidential Computing: architecture, security, attestation, configuration and deployment of SGX enclaves, confidential VMs/GPUs, AKS containers and Fortanix CCM. CX Codex +1 174installs chainlink-data-feeds hedera-dev chainlink-data-feeds helps safely read Chainlink AggregatorV3 price feeds on Hedera testnet (296) and mainnet (295), validate round staleness, and normalize decimals. CX Codex +1 577installs azure-confidential-ledger MicrosoftDocs A skill for Claude Code, Codex, and Cursor that guides Azure Confidential Ledger development. Supplies architectures, quota checks, and deployment templates directly into agent context. CX Codex +1 166installs Synmerco — Confidential Escrow for AI Agent Commerce joelasota Synmerco — Confidential Escrow for AI Agent Commerce is a Claude Code MCP server for agent-to-agent commerce with escrow, trust scores, and delivery proofs for confidential work products. CX Codex +1 3,239installs use-link-identity stripe use-link-identity is a Claude Code skill for retrieving Link attestations and presenting selected signed claims to a service. It activates when a service asks for an AAT or identity proof. CX Codex +1 15,700installs

From the same set7

Project sponsors

Companies that keep this analytics open