Cursor Skills for Web3 Developers
Cursor skills for Web3 developers are SKILL.md files in .cursor/skills/ that teach Cursor's Agent the chain-specific steps in your project, such as checking the network and signer setup before touching a wallet flow, or running the contract tests before and after a Solidity change.
Last updated By Gizmolab
Why generic prompts break down in Web3 repositories
A Web3 codebase mixes wallet code in the frontend, contract calls, RPC and chain configuration, SDK wrappers and security-sensitive logic. A one-off prompt rarely carries all of that, so the Agent fills the gaps with generic patterns: the wrong chain id, a made-up SDK method, or a fix that hides a revert instead of explaining it.
A skill writes those checks down once. The Agent reads the real configuration first, runs the project's own tests, and stops before anything touches mainnet or a key.
Four Web3 skills you can copy
Save each file at the path in its caption. Change folder names, test commands and networks to match your project. New to skills? Start with How to create a Cursor skill.
Debug a wallet flow
Use it when: A wallet will not connect, a signature or transaction fails, or the app shows the wrong network or balance.
---
name: wallet-flow-debug
description: Debugs wallet connection, signing, chain switching and transaction submission in a dapp frontend. Use when a wallet will not connect, a signature or transaction fails, or the app shows the wrong network or balance.
---
# Debug a wallet flow
## Steps
1. Identify the wallet stack from package.json (for example wagmi and viem, ethers, or @solana/wallet-adapter) and read how the provider is configured.
2. Find the chain configuration: chain ids, RPC URLs and the environment variables that set them.
3. Trace the failing flow in order: connect, check the chain, sign or send, wait for confirmation, update the UI.
4. Quote the exact error. A user rejection, the wrong chain, insufficient funds and a contract revert each need a different fix.
5. Propose the smallest fix and say how to test it on a testnet.
## Limits
- Never ask for, log or hard-code a private key or seed phrase.
- Do not switch chains or RPC providers as a fix unless the user asks.
- Test on a testnet, not mainnet.Review a Solidity change
Use it when: Contracts changed and you want a security and correctness pass before merging or deploying.
---
name: solidity-change-review
description: Reviews a change to Solidity contracts for security and correctness before it is merged or deployed. Use when contracts change or the user asks for a contract review.
paths:
- "contracts/**/*.sol"
- "src/**/*.sol"
---
# Review a Solidity change
## Steps
1. Read the diff and every function it touches, including inherited contracts and modifiers.
2. Run the tests: `forge test` for Foundry, or `npx hardhat test` for Hardhat.
3. Check each changed function for:
- access control on every state-changing path
- external calls made before state updates (reentrancy)
- unchecked return values and unsafe casts
- events for state changes that off-chain code relies on
- storage layout changes, if the contract is upgradeable
4. Note any gas increase in functions users call often.
## Output
Findings grouped as Critical, High, Medium and Low, each with the file, the function, the problem and a fix. End with the test command you ran and its result.
## Limits
- Review only. Do not edit contracts unless the user asks.
- This is not an audit. Flag anything that needs a professional review.Deploy to a testnet
Use it when: Deploying with the project's own script. Manual-only, so the Agent never starts a deploy by itself.
---
name: testnet-deploy
description: Deploys the contracts to a testnet with the project's own deploy script and records the addresses. Runs only when the user types /testnet-deploy.
disable-model-invocation: true
---
# Testnet deploy
## Before deploying
1. Confirm the target network with the user and check that it is a testnet, such as Sepolia or Base Sepolia.
2. Run the full test suite. Stop if anything fails.
3. Check that the RPC URL and the deployer account come from environment variables or an encrypted keystore, never from a file in the repo.
## Deploy
4. Run the deploy script the project already has. For Foundry this is usually `forge script script/Deploy.s.sol --rpc-url <testnet> --broadcast` plus your account flag. Do not write a new script.
5. Record each deployed address and transaction hash in the deployments file the project uses.
## Limits
- Never deploy to mainnet from this skill.
- Never print, store or commit a private key.Change an Anchor program
Use it when: Editing a Solana program built with Anchor: instructions, accounts, PDAs and the IDL clients depend on.
---
name: anchor-program-change
description: Makes and checks a change to a Solana program built with Anchor. Use when editing files under programs/ or when the user asks to change an instruction, an account or a PDA.
paths:
- "programs/**/*.rs"
---
# Change an Anchor program
## Steps
1. Read the instruction handler and its accounts struct before editing.
2. Check every account constraint: signer, mut, has_one, seeds and bump, owner.
3. If the change adds or reorders fields in an account, say how existing accounts will be migrated or resized.
4. Run `anchor build`, then `anchor test`.
5. If the IDL changed, list the client code that must be regenerated or updated.
## Limits
- Do not change the program id or the PDA seeds of a deployed program without asking.
- Test on a local validator or devnet, never mainnet.Keep keys and funds out of reach
- Never put a private key, seed phrase or API key in a SKILL.md or a rule. Skills are plain text, and project skills are usually committed.
- Read secrets from environment variables or an encrypted keystore.
- Make anything that sends transactions manual-only with disable-model-invocation: true.
- Keep deploy skills on testnets, and review every transaction before you sign it.
- Read a third-party skill before adding it. The Agent follows its instructions and may run its scripts.
Gizmolab's skills on ClawHub
Alchemy Web3
Lets the Agent query balances, tokens, NFTs, transfers and blocks through Alchemy's APIs, which pairs well with the wallet debugging skill above. Needs your own Alchemy API key, read from an environment variable.
View on ClawHubGizmoLab Tools
Tells the Agent which free tool on this site fits a task, such as decoding a transaction, calling a contract or creating a Solana token.
View on ClawHubBoth use the SKILL.md format and were written with OpenClaw in mind. Install the Alchemy skill into your project with npx skills add 0xGizmolab/alchemy-web3-skill -a cursor, which saves it in .agents/skills/. For GizmoLab Tools, copy its files from ClawHub into .cursor/skills/<skill-name>/. More on how skill folders work is in Cursor skills.
Free tools to use alongside these skills
Paste a failing transaction's hash or input data and see what it actually called.
Load a verified contract and call its functions, to check what the Agent changed.
Turn a failing fuzzer trace into a Foundry test the review skill can run.
Generate Foundry, Echidna or Medusa invariant tests and handlers from a contract's ABI.
Free devnet SOL for Solana wallet tests and for trying an Anchor program on devnet.
Work in Claude Code as well? The same SKILL.md files work there from .claude/skills/, and Cursor reads that folder too. See Claude Code skills.
FAQ
What makes a good Cursor skill for Web3 developers?
Should a Cursor skill ever contain a private key or seed phrase?
Should Cursor be allowed to deploy contracts on its own?
Can one skill cover both Solidity and Solana projects?
Which Gizmolab skills can I use in Cursor?
Related guides
- Cursor skills
What a Cursor skill is, where Cursor looks for SKILL.md files, and how the Agent uses them.
- How to create a Cursor skill
Build a skill step by step, from the folder and frontmatter to testing it in Agent chat.
- Cursor skill examples
Complete SKILL.md files for edits, debugging, refactors, reviews and planning.
- Cursor AI rules
Project rules in .cursor/rules, the four rule types, .mdc templates and AGENTS.md.
- How to reload Cursor skills
What to check when a skill or rule does not load, and how to reload Cursor safely.
Using a different tool? See Claude Code skills, OpenClaw skills or the AI skills hub.
Sources
This guide was checked against the official documentation on 5 November 2026. Cursor changes quickly, so when something here differs from the docs, the docs win.