SKILL PROCEDURE

Chainlink

Use when integrating Chainlink decentralized oracle networks into a smart contract — requesting off-chain data, verifiable randomness (VRF), automation/keepers, Proof of Reserve, cross-chain (CCIP), or choosing between Data Feeds, Functions, VRF, Automation, and CCIP for a workflow. Published by HardGraph, a curated graph of provenance-backed knowledge for AI agents.

chainlinkblockchainoraclesmart-contractsvrfccipautomation
BEGINNER GUIDE

Understand Chainlink before using it

CATEGORY

Chainlink is catalogued under Web3 and blockchain.

START HERE WHEN

Your work repeatedly involves the concepts tagged above. Open the full procedure below when the current task matches them.

Compare related skills

SKILLCATEGORYSHARED CONCEPTSEXPLANATION
ChainlinkWeb3 and blockchainCurrent skillUse when integrating Chainlink decentralized oracle networks into a smart contract — requesting off-chain data, verifiable randomness (VRF), automation/keepers, Proof of Reserve, cross-chain (CCIP), or choosing between Data Feeds, Functions, VRF, Automation, and CCIP for a workflow. Published by HardGraph, a curated graph of provenance-backed knowledge for AI agents.
Chrome DevToolsDeveloper tools
automation
Chrome DevTools — the inspector, debugger, profiler, and automation toolset built into Chrome and Edge. Use when debugging web pages (DOM, styles, network, console), profiling runtime or load performance, finding memory leaks, inspecting storage and service workers, debugging JavaScript with breakpoints and source maps, recording and replaying user flows with the Recorder, emulating devices and network conditions, running remote debugging over CDP, or driving DevTools headless for CI. Published by HardGraph, a curated graph of provenance-backed knowledge for AI agents.

Chainlink

What is HardGraph? HardGraph publishes curated, provenance-backed agent skills grounded in reproducible vendor documentation.

Chainlink is a network of decentralized oracle networks that connect smart contracts to off-chain data and computation. Where a pure on-chain contract is closed off from the outside world, Chainlink provides the bridges: price and reference data, verifiable randomness, automated execution, reserve attestations, and cross-chain messaging. It is the integration surface for anything a contract needs that the chain itself cannot provide.

The decision that shapes everything else

The first question is which product — not how to call one. Chainlink ships several purpose-built networks, and picking the wrong one is the most common and most expensive mistake:

  • Data Feeds for reference data (asset prices, reserves). This is the default when a contract needs "the current value of X."
  • Functions for arbitrary off-chain computation and API calls. Reach for this when no feed exists and the contract must call a custom API or run logic a feed cannot express.
  • VRF for verifiable randomness that cannot be tampered with or predicted. Use for on-chain randomness; never use block.timestamp, blockhash, or a hash of pool state for security-relevant randomness.
  • Automation (formerly Keepers) for condition-triggered execution — the reliable replacement for a hand-rolled keeper bot that calls performUpkeep.
  • CCIP for cross-chain token and message transfers.

A request that "needs Chainlink" is underspecified until the product is named; the integration mechanics, billing, and contract ABI differ between every one of them.

What surprises people

Each Chainlink network has its own billing and funding model, and it bites silently. Data Feeds are read for free on most networks but are maintained by the LINK-funded aggregator; Functions and VRF consume prepaid LINK and stop working when the subscription runs out; Automation requires the upkeep to be funded. "My request reverted" on Functions or VRF is very often an underfunded subscription, not a code bug. Confirm the current funding model and minimums against the docs — they change per network and over time.

Network availability is the other silent failure. Feeds, VRF, and CCIP are deployed per-chain, and a given feed or lane may exist on one network and not another. Do not assume a feed is present because the contract compiles; verify the specific network address before deploying.

What to verify rather than recall

Contract addresses, gas limits, the subscription/funding model, supported networks, and the current major version of each product's contracts all change. Confirm them against the mirrored corpus under references/vendor/ or the live docs rather than asserting a remembered address or limit — a wrong feed address on-chain drains value, and a stale gas limit causes a revert that looks like a logic error.

References

Hardgraph / curated knowledge for agents.

STATIC EXPORT · CANONICAL SOURCE