OPEN-SOURCE AGENT SKILL / XRP LEDGER

Explore XRPL.Keep the keys.

XRPL Muse helps terminal-based agents inspect the ledger and prepare transaction proposals—while approval and signing stay inside a deliberate human-controlled boundary.

Open the source on GitHub ↗Ask Terramike AI about Muse ✦

The control boundary

Human in the loop by design

Propose → review → approve → sign.

XRPL Muse separates preparation from authorization. The agent may help assemble and explain an action, but an exact proposal must still pass human review before the protected signing path can proceed.

01 · ProposePrepare the intended XRPL action and display its details.

02 · ReviewCheck accounts, assets, amounts, fees, limits, expiry, and network.

03 · ApproveA human explicitly approves the exact proposal hash.

04 · SignThe separate signer re-checks policy before submission.

What it helps with

Start on testnet

01 INSPECT

Read the ledger.

Explore balances, offers, trust lines, NFTs, marketplace information, and other read-only XRPL state.

02 PREPARE

Build proposals.

Prepare payments, offers, trust lines, cancellations, and selected NFT actions for human review.

03 EXPLAIN

Show the ceremony.

Make network, accounts, assets, limits, fees, expiry, and proposal identity visible before approval.

Start with the source

Python 3.10+ · testnet first

04 QUICK STARTCLONE LOCALLY

Practice before value is at risk.

Review the repository and current documentation, then begin with an isolated testnet setup.

git clone https://github.com/terramike/xrpl-muse-skill.git
cd xrpl-muse-skill
./install.sh

05 DOCUMENTATION

Read before running.

The wiki covers setup, command behavior, policy controls, security boundaries, and troubleshooting.

Open the wiki ↗

What the upstream proof means

Scope matters

06 VERIFIED SCOPE

The command source compiled and its fixture behavior passed.

The merged HOL-guard extension classifies Muse commands so proposal, approval, and signing-gate transitions can require review while recognized read-only and safe help/list variants remain inert. The external trust class remains off until locally enabled.

! NOT A BLANKET CERTIFICATION

Evidence, not magic.

This does not certify every operating system, wallet, signer configuration, future transaction, or user decision. Protected secrets, limited wallets, human review, and testnet-first practice still matter.

Plain-language answers

Frequently asked

Is XRPL Muse an autonomous trading bot?

No. It has no scheduler, market watcher, or automatic approval flow. Its defining pattern is proposal first, followed by explicit human review and approval.

Does the agent receive my signing secret?

The intended design keeps signing secrets outside the proposal agent. Never paste a seed or secret into chat, a proposal, source control, or ordinary configuration.

What did HOL-guard test?

The merged contribution defined seven permissions, seven rules, and 21 portable behavior cases for command classification. All 21 passed with zero target commands executed during the fixture.

Should I start on XRPL mainnet?

No. Start on testnet, use a dedicated setup, read the current security documentation, and understand every proposal before considering real value.

BUILD WITH CURIOSITY · OPERATE WITH CONTROL

See the skill. Read the boundary.

XRPL Muse is public, inspectable, and designed to keep consequential approval with the human using it.

Explore XRPL Muse on GitHub ↗Review the upstream evidence ↗