Chainlink Developer Agent Skills
Nine skills for building on Chainlink: CRE, CCIP, Data Feeds, Data Streams, VRF and other network products
- 128 ★ on the repository
- Last commit
Extended access
Why this level
- Skills write contracts and scripts that can move tokens across chains, and the code is not audited
- Working with the CLI and SDKs needs wallet and node keys, which must be kept outside the repository
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/chainlink-developer-agent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Skills write contracts and scripts that can move tokens across chains, and the code is not audited
- Working with the CLI and SDKs needs wallet and node keys, which must be kept outside the repository
This is third-party code. Review the repository files before installing.
What it does
A repository from the smartcontractkit organization with nine skills following the Agent Skills standard. Each skill covers one Chainlink product: chainlink-cre-skill (onboarding, workflow generation, CLI and SDK), chainlink-cre-connect-skill (event watchers, gasless operations, Smart Accounts), chainlink-ccip-skill (cross-chain messages, contracts, local tests, monitoring), chainlink-data-feeds-skill (reading price feeds and multi-chain integration), chainlink-data-streams-skill (REST and WebSocket SDKs, report decoding, on-chain verification), chainlink-vrf-skill (VRF v2.5 consumers, migration from V2, billing), chainlink-ace-skill and chainlink-confidential-ai-attester-skill. chainlink-nop-skill diagnoses a single CCIP and CRE node operator: it makes only GET requests driven by a local YAML config and investigates against source code. Skills keep long reference material in `references/`, and the needed pieces load by request triggers instead of being read all at once. The author recommends invoking skills explicitly. The README shows a prompt where three skills together build a tokenized fund project: on-chain asset data, token issuance, reserve verification and cross-chain transfers.
Who it is for. Smart contract and backend developers building on Chainlink who want the agent to know the current interfaces of the network products.
Good fit when
- A contract needs an asset price from a Data Feeds feed and it is unclear how that differs from a Data Streams report
- A token has to move between chains through CCIP with a local test and later transfer monitoring
- Random numbers for a game or a draw come from VRF v2.5 while the old code still runs on V2
- CRE workflows need generation, simulation and CLI error triage
- A node operator is working out why CCIP or CRE behaves oddly and wants diagnostics that never write to the node
Not a fit when
- You need a production system without an audit: the repository labels the code as examples for learning and disclaims responsibility for errors
- The task is outside Chainlink, for example a pure DeFi protocol without oracles: the nine skills do not cover it
- You need support from Chainlink Labs: the README states outright that the repository is not a Chainlink product or service
- The agent does not support Agent Skills: without that it will not pick the skills up
Example request
Using /chainlink-data-feeds-skill, read the ETH/USD price in a Foundry contract and add a test that checks data freshnessLimitations
The README disclaimer says the repository holds community examples and is no Chainlink product, and neither Chainlink Labs nor the Chainlink Foundation supports it. The code may not be audited, so key management, testing and review before mainnet stay with you. It falls outside the Chainlink bug bounty programs. The first line of the README still calls the repository official. GitHub could not detect the license automatically.
How to disable. Remove skills with `npx skills remove` and the skill name, or delete the `chainlink-*-skill` folders by hand from the project `.agents/skills` or the global skills directory.
FAQ
How do I install the Chainlink skills in Claude Code?
Run `npx skills add smartcontractkit/chainlink-agent-skills` in the project root. The `-g` flag installs globally and `--skill` picks one, for example `chainlink-vrf-skill`.
How do Data Feeds differ from Data Streams?
Data Feeds read a price from a contract through `AggregatorV3Interface` and `latestRoundData`, while Data Streams deliver reports over REST or WebSocket for on-chain verification. The Data Feeds skill sends stream questions straight to `chainlink-data-streams-skill`.
Do I have to call a skill manually?
No, but the author recommends calling it explicitly, for example `/chainlink-ccip-skill`. An agent with Agent Skills support picks skills up from the task description on its own.
What does chainlink-nop-skill do?
It diagnoses a single CCIP and CRE node operator, sending only GET requests driven by a local YAML config. The README per-skill install list omits it, but it is in the table and ships with the all-in-one install.
Which Chainlink products do the skills cover?
Nine skills: CRE, CRE Connect, CCIP, node operator diagnostics, Data Feeds, Data Streams, ACE, Confidential AI Attester and VRF. The right skill loads based on the topic of the request.
Discussion
No questions or reviews yet.
Related
SkillsTech Leads Club Agent Skills
Agent Skills
A vetted skill registry with a CLI that installs into twenty-plus agents and an MCP server for on-demand skill lookup
A set of independent skills against drift in large codebases: hard-cut, root-cause tracing, lean testing and CI right-sizing.
Over 1,100 ready-made skills for Claude Code, Codex, Gemini CLI and Cursor, plus an MCP server that picks a skill for the task
A library of 30-plus instruction skills for coding agents: planning, building, testing, hallucination guards and shipping
Nine personal engineering skills: an idea-to-PR chain, standards search, a sandboxed review loop and wizard pages for changes
A Tencent CLI and two skills that repair broken agent skills: they capture a failure, make a tested fix and open a local PR.
README in short
The README opens with a table of nine skills and their areas, followed by project and global install commands, a large example prompt using three skills and a long disclaimer that these are community examples.
README badge
Original SKILL.md
--- name: chainlink-data-feeds-skill description: "Helps developers read Chainlink price feeds with AggregatorV3Interface and latestRoundData. This is not a Data Streams report. REST and WebSocket report streams go to chainlink-data-streams-skill. It applies to price feed integration, feed address lookup, consumer contracts, multi-chain feeds, MVR bundle feeds, SVR and OEV feeds, feed monitoring, historical data, L2 sequencer checks, rates and volatility feeds, and SmartData and RWA feeds." license: MIT allowed-tools: Read WebFetch Write Edit metadata: version: "0.0.9" --- # Chainlink Data Feeds Skill ## Progressive Disclosure | Request / triggers | Load | |---|---| | EVM read, consumer contract, off-chain read, address lookup/debugging | [price feeds](references/reading-price-feeds.md) | | Working Foundry/Data Feeds project | Load the [starter README](templates/starter-kit/README.md) and every listed file; reply with their contents or a tailored derivative, never a local/absolute path alone. Preserve layout and include install/test/deploy commands, Sepolia ETH/USD or verified named-network lookup, unaudited/not-for-production warning, and L2 sequencer protection. | | Multiple-Variable Response, multi-variable feed, `BundleAggregatorProxy` | [MVR](references/mvr-feeds.md) | | Smart Value Recapture, OEV/MEV recapture, searcher onboarding | [SVR](references/svr-feeds.md) | | Feed categories/selection, SmartData/RWA, rates/volatility, tokenized equity | [types](references/feed-types.md) | | Solana, StarkNet, Aptos, Tron, Move, Cairo, Anchor, TronBox | [multi-chain](references/multi-chain.md) | | L2 sequencer, deprecation, monitoring, registry, responsibilities, data sources, self-managed feeds | [operations](references/feed-operations.md) | | Live addresses/schedules/parameters; interface/signature mismatch, proxy/aggregator source, GitHub example | [official sources](references/official-sources.md), then [docs index](assets/data-feeds-docs-index.md) for facts the curated references do not have | Do not load references speculatively. Default EVM requests to price feeds; ask only when routing depends on missing context. Do not assume this skill is the only capability available. Hand CCIP/bridge, VRF, Automation, and Functions requests to their owning products exclusively, giving each product's standard next steps from general knowledge; emit no Data Feeds or weaker substitute and never improvise live constants. When a user requests Data Feeds code or operational steps, emit that deliverable in the response; never replace it with a status or completion summary. For MVR requests, choose the MVR route even when a live address, feed ID, or schema cannot be verified; use clearly marked placeholders for unverified values and continue with all requested design, consumer code, tests, and configuration instead of stopping at verification. Every SVR response must emit the exact standalone sentence `Complete the Chainlink compatibility form` before any onboarding or address-migration guidance. For a configuration-focused SVR protocol integration or migration, give concise configuration, onboarding, verification, and safety steps plus only the minimal address/configuration delta; do not emit a full consumer unless the user explicitly requests one. Keep auction mechanics high-level and exclude searcher-only selectors, WebSocket endpoints, signing instructions, and MEV-Share/Atlas bot implementation details unless the user explicitly asks to build or operate a searcher. ## Safety rules 1. Validate `updatedAt` against the feed heartbeat; reject incomplete or stale rounds. 2. Call `decimals()` at runtime; never hardcode decimals. 3. On L2, always use `AggregatorV2V3Interface` for the sequencer uptime feed—never `AggregatorV3Interface`; reject `answer != 0` and enforce the 3600-second recovery grace period. 4. Never use deprecated `answeredInRound` for freshness or other logic. 5. For an implicit current-price Solidity getter, use the exact named-return form `returns (int256 price, uint8 decimals)`; declare `uint256 updatedAt;` first, then assign `(, price,, updatedAt,) = dataFeed.latestRoundData();`. Do not substitute an unnamed-return or alternate compiling tuple form. 6. Warn that examples are unaudited and not for production; mainnet integrators own audits, dependencies, monitoring, circuit breakers, contingencies, and user risk disclosure. ## Freshness policy 1. Use this skill's references first. 2. If [official-sources.md](references/official-sources.md) names the page, fetch that page's `.md` URL.