wallet skill for Starchild agents: balances, transfers and signing on EVM and Solana

The Starchild team's wallet skill: lets a Starchild agent check balances, send tokens and sign data on EVM chains and Solana.

Installs: 8.9K; on skills.sh
Stars: 29; on GitHub
Files: 4; besides SKILL.md
Updated: Oct 2026; last repo commit

wallet is a Python script skill for Starchild agents. Functions check EVM and Solana balances, broadcast or sign transactions, sign messages and typed data, read transaction history and check or propose a wallet policy. Transfers are real: the skill says the policy default is off, which allows everything, and a confirmation card appears only once a policy is enabled. It runs inside the Starchild platform, not in a plain Claude Code setup.

Good for

Users of Starchild agents who want the agent to hold and move funds
Developers calling wallet functions from Starchild task scripts
People who want to set a wallet policy that limits what the agent can do

Install wallet

Any client

Run this in a terminal.

shell

npx skills add Starchild-ai-agent/official-skills --skill wallet

Requires Node.js.

The README describes these skills as maintained by the Starchild team and installed with the skills CLI. The code imports modules from the Starchild agent platform, so it is built to run there. This page has no Claude Code or Claude app tab for that reason.

Built from the package data we hold. Not yet tested by us. Source.

What's in the skill

When it loads

Multi-chain wallet: EVM and Solana balances, transfers, signing, and policy. Use when checking balances, sending tokens, signing typed data, or proposing a wallet policy (e.g. send 10 USDC on Base, sign EIP-712, Solana balance).

Sections in SKILL.md

How to call
Functions (from core.skill_tools import wallet)
The User's Own Wallets (login / secondary)
Key Facts
Workflows
Check balances
Send a transaction (EVM)
Sign EIP-712 typed data
Policy Management
Standard Wildcard Policy (when needed)
Policy Modes — CRITICAL DECISION TABLE
Mode 1: Allow-All (Standard Wildcard)
Mode 2: Deny-All (Lockdown)
Mode 3: Whitelist (Selective Allow)
Privy Policy Rules — Key Constraints
Gotchas

Key facts

Author: The Starchild team (README: official skills maintained by the Starchild team)
License: No LICENSE file at the repo root or in this skill's folder
Installs on skills.sh: 8,860 as of 2026-11-02
GitHub stars: 29 for the whole official-skills repo, 2026-11-02
What is inside: 5 files including SKILL.md: __init__.py, exports.py, wallet.py and logo.png
Version: 3.7.0 in the SKILL.md frontmatter
Runs: Python functions called with python3 -c from bash, from core.skill_tools.wallet
Chains: All DeBank-supported EVM chains, plus Solana
Last repo commit: 2026-10-30
Works in: Starchild agents. The README documents the skills CLI and search_skills inside a Starchild conversation, and the code imports Starchild platform modules.

What it includes

Lists the agent's wallet addresses and your own linked wallets
wallet_info returns the agent wallet addresses. get_user_wallets reads your login and secondary wallet addresses from environment variables the platform injects. SKILL.md calls the second one read-only and says the agent holds no keys for those wallets.
Try: “What is my login wallet address?”
Read · wallet_info and get_user_wallets
wallet_balance, wallet_sol_balance, wallet_get_all_balances
EVM balances come from DeBank and Solana balances from Birdeye. wallet_balance needs a chain name, and wallet_get_all_balances checks every chain at once.
Try: “Check my balance on Base.”
Network · Balances
wallet_transfer and wallet_sol_transfer broadcast real transactions
EVM amounts are in wei. Gas is sponsored by default and falls back to your balance. SKILL.md says to check the balance first and the history afterwards.
Try: “Send 0.01 ETH on Base to 0x...”
Write · Transfers
Sign transactions without broadcasting, sign messages and typed data
wallet_sign_transaction and wallet_sol_sign_transaction sign without sending. wallet_sign does EIP-191 messages, wallet_sign_typed_data does EIP-712, and wallet_sol_sign signs Solana messages.
Write · Signing
wallet_transactions and wallet_sol_transactions
Lists recent EVM and Solana transactions for a chain and asset.
Network · History
wallet_get_policy and validate_and_clean_rules, then a UI card to propose
Checking and validating are script functions. Proposing a policy goes through the frontend_action tool, and you confirm and sign in the UI. Call it once per chain.
Write · Policy
Allow-all, deny-all lockdown and whitelist
SKILL.md notes that DENY beats ALLOW. Allow-all denies key export and allows everything else. Whitelist allows only listed operations, such as transfers to one address.
Read · Policy modes

From the official docs

Permissions

High risk

Claude can send real crypto and sign data from the agent's wallet, and by default no confirmation is asked. Fund it with only what you can lose, and set a whitelist policy before use.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It can send real funds
wallet_transfer and wallet_sol_transfer broadcast transactions. SKILL.md says the wallet policy defaults to off (allow-all) and only an enabled policy makes transactions need UI confirmation; its send workflow lists check balance, transfer, then verify, with no approval step. Our advice: do not run this skill unattended. Fund the agent wallet with only what you can lose, and set a whitelist policy first. github.com
It signs arbitrary typed data
wallet_sign_typed_data signs EIP-712 data, and SKILL.md's example is a Permit. A signed permit can authorize token spending, so check what is being signed. github.com
The standard policy allows everything
The standard wildcard policy in SKILL.md denies key export and allows every other method. A whitelist policy that allows only named operations is stricter. github.com
Balance lookups call third-party APIs
exports.py and wallet.py call DeBank and Birdeye. SKILL.md says the DeBank and Birdeye keys are injected by sc-proxy, so you do not set them. github.com

Technical details

Turning on read-only	No read-only mode. Only the balance, history and policy-check functions are reads; the transfer and signing functions are in the same skill.
Talks to	https://pro-openapi.debank.com (EVM balances). https://public-api.birdeye.so (Solana balances). A wallet service for transfers and signing (endpoints such as /agent/transfer)
Sign-in and keys	The code reads BIRDEYE_API_KEY and DEBANK_API_KEY from the environment, and SKILL.md says the platform injects them. Transfers and signing are sent to a wallet service, for example through /agent/transfer.
Data retention	SKILL.md names no logging or telemetry.
Scopes	Runs Python functions through bash that can broadcast EVM and Solana transactions from the agent's wallet Signs messages and typed data with the agent's wallet Reads your login and secondary wallet addresses from environment variables

Questions

What does the wallet skill do?

It gives a Starchild agent an EVM and Solana wallet: check balances, send tokens, sign messages and typed data, read transaction history, and set a policy that limits what the agent may do.

Can I use it in Claude Code?

Not as written. The code imports Starchild platform modules such as core.skill_tools and tools.wallet, and the repo is a set of Starchild skills. Use it in a Starchild agent.

How do I install wallet?

Use the install box at the top of this page: the skills CLI, or search_skills inside a Starchild conversation.

Is wallet safe to use?

It can broadcast real transactions and sign data from the agent's wallet, so treat it as high risk. SKILL.md says the policy is off by default (allow-all) and that only an enabled policy makes transactions need UI confirmation. The send workflow in SKILL.md checks the balance first and the history afterwards, with no approval step. Our advice: do not run it unattended. Fund the agent wallet with only what you can lose, enable a whitelist policy first, and read each transfer before it runs.

Does it ask before sending funds?

SKILL.md gives no confirmation step for script transfers. Only when a policy is enabled do transactions need UI confirmation. Transactions from your own wallet go through a UI signing flow instead.

Does it need API keys?

The code reads BIRDEYE_API_KEY and DEBANK_API_KEY from the environment. SKILL.md says sc-proxy injects them, so on the Starchild platform you do not set them.

Is it official?

Yes. It is in the Starchild team's official-skills repo, and the README describes the skills as maintained by the Starchild team.

What is the difference between the agent wallet and my wallet?

SKILL.md says they are different. The agent wallet is the one the agent controls. Your own login and secondary wallets are read-only to the agent, and transactions from them go through a UI signing flow.

Information

updated Nov 2, 2026

Repository: starchild-ai-agent/official-skills
Folder: wallet

starchild-ai-agent/official-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 6

README · github.com
CONTRIBUTING.md · github.com
wallet on skills.sh · skills.sh
wallet SKILL.md · github.com
wallet/exports.py · github.com
wallet/wallet.py · github.com