https://github.com/smartcontractkit/chainlink-agent-skills
[PUBLIC] Repository for Chainlink Skills that implement https://agentskills.io/specification
https://github.com/smartcontractkit/chainlink-agent-skills
Last synced: about 1 month ago
JSON representation
[PUBLIC] Repository for Chainlink Skills that implement https://agentskills.io/specification
- Host: GitHub
- URL: https://github.com/smartcontractkit/chainlink-agent-skills
- Owner: smartcontractkit
- License: mit
- Created: 2026-02-11T00:38:46.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-04-28T16:25:09.000Z (about 2 months ago)
- Last Synced: 2026-04-28T18:21:55.262Z (about 2 months ago)
- Language: JavaScript
- Homepage:
- Size: 457 KB
- Stars: 94
- Watchers: 3
- Forks: 10
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Codeowners: .github/CODEOWNERS
- Agents: AGENTS.md
Awesome Lists containing this project
- awesome-hermes - chainlink-agent-skills
- awesome-hermes-agent - production
- awesome-hermes-agent - chainlink-agent-skills
- awesome-crypto-skills - Chainlink: agent-skills - Chainlink CRE onboarding, CCIP cross-chain operations (Blockchain Operations)
README
# Chainlink Agent Skills
Official Repo for Chainlink skills. Each skill follows the [Agent Skills specification](https://agentskills.io/specification).
## Available Skills
| Skill | Description |
| ------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
| [chainlink-cre-skill](chainlink-cre-skill/) | CRE onboarding, workflow generation, CLI/SDK help, and runtime operations |
| [chainlink-ccip-skill](chainlink-ccip-skill/) | CCIP sends, contracts, local testing, monitoring, discovery, and CCT workflows |
| [chainlink-data-feeds-skill](chainlink-data-feeds-skill/) | Data Feeds contracts, multi-chain Data Feeds integration |
| [chainlink-data-streams-skill](chainlink-data-streams-skill/) | Data Streams REST/WebSocket SDKs, report decoding, on-chain verification, and real-time frontend apps |
| [chainlink-ace-skill](chainlink-ace-skill/) | ACE core contracts, Policy Management, Cross-Chain Identity, and compliance token examples |
## Install
Use [vercel's CLI for the open skills ecosystem](https://github.com/vercel-labs/skills#readme). Project-level installation is the default.
```bash
npx skills add smartcontractkit/chainlink-agent-skills
```
But if you want to install globally (at the user level) then add the `-g` flag.
Note the use of `--skill` to specify which specific skill to install.
```bash
npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-cre-skill -g
npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-ccip-skill -g
npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-data-feeds-skill -g
npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-data-streams-skill -g
npx skills add smartcontractkit/chainlink-agent-skills --skill chainlink-ace-skill -g
```
## Use
When your agent supports Agent Skills, it will discover and activate these skills based on the task. **However** we recommend that you explicitly invoke the skill in your agent chat sessions as follows:
```text
Using the /chainlink-data-feeds-skill, develop a Solidity contract that:
- reads the ETH/USD Chainlink price feed on Ethereum mainnet
- is structured as a Foundry project
- handles required package installation and project setup
- includes proper validation
- follows security best practices
Any questions before you start?
```
## Evaluate And Improve Skills
This repo includes two local eval paths:
- `evals/run-agent-eval.md` runs smoke/full evals through agent subagents without API keys.
- `evals/run-agent-ab-test.md` compares baseline responses against skill-enabled responses using subagents, then aggregates wins, ties, regressions, and recommended skill changes.
For example:
```text
Run an agent A/B test for chainlink-cre-skill using mixed-chainlink
Improve chainlink-cre-skill using agent A/B tests
```
The A/B workflow is useful in Cursor, Codex, Gemini CLI, and similar tools because it spends local agent/subagent budget instead of external model API keys.