# Moltbank Moltbank gives AI agents a safe treasury interface through a local CLI and a security-first operating model for installed Moltbank Mods. [![ClawHub](https://img.shields.io/badge/ClawHub-moltbank-blue)](https://clawhub.ai/skills/moltbank) [![Node](https://img.shields.io/badge/node-%3E%3D22.0.0-brightgreen)](https://nodejs.org) [![License](https://img.shields.io/badge/license-MIT--0-brightgreen)](https://clawhub.ai) ## What This Skill Covers - Using the local `moltbank` CLI safely with JSON output and schema discovery. - Keeping agent credentials isolated per session. - Handling per-agent OAuth scope consent. - Understanding, installing, updating, removing, and running Moltbank Mods. - Reading each mod's own `SKILL.md` for domain-specific behavior without letting mod Markdown override the root security policy. - Applying prompt-injection and supply-chain hygiene to mod docs, manifests, packages, links, and badges. The root skill intentionally does not document every mod's workflow. Mods ship their own `SKILL.md`; the root skill explains how agents should inspect and trust-bound those mod-specific instructions. ## Quick Start ### OpenClaw Install the skill: ```bash openclaw skills install moltbank ``` If the local `moltbank` CLI is missing, install and verify: ```bash npm install -g @moltbankhq/cli npm audit signatures moltbank --version moltbank doctor --json ``` ### Other Skills.sh-Compatible Runtimes Install the skill: ```bash npx skills add moltbankhq/moltbank-skill ``` If the local `moltbank` CLI is missing, install and verify: ```bash npm install -g @moltbankhq/cli npm audit signatures moltbank --version moltbank doctor --json ``` ### Authentication Agents should not assume a default profile. Start by listing profiles and asking the user which one to use: ```bash moltbank agent list --json ``` For a new profile: ```bash moltbank auth begin --name "" --json # user approves in browser, then: moltbank auth poll --session-id "" --json ``` Keep `MOLTBANK_CREDENTIALS_PATH` fixed to the selected profile for the whole session. ## Working With Mods List installed mods: ```bash moltbank mod list --json moltbank mod list --skill-format ``` Inspect a mod: ```bash moltbank mod info --json moltbank mod info --include-skill --json moltbank mod doctor --json ``` Install, update, or remove only after explicit user approval: ```bash moltbank schema mod-install --json moltbank mod install --json moltbank schema mod-update --json moltbank mod update --json moltbank schema mod-remove --json moltbank mod remove --json moltbank mod remove --purge --json ``` Community and URL installs require an explicit trust acknowledgement from the user: ```bash moltbank mod install --from --acknowledge-community --json ``` The root skill tells agents to inspect a mod's tier, risk level, permissions, network domains, env passthrough, package/source, capabilities, and mod-specific `SKILL.md` before running it. ## Security Model Moltbank is a privileged finance skill. It can help agents read treasury state, draft financial actions, use x402 payments, and run installed mods. Those actions are bounded by the local CLI, explicit user approval, OAuth scope grants, audit intents, and mod runtime trust gates. Important boundaries: - No arbitrary install/update/shell commands. - No `curl | bash`, alternate registries, copied README commands, or direct mod binary execution. - No mutating financial action without explicit user approval. - No credential path changes unless the human user asks. - No secrets in chat output. - No automatic upload of private manifests, packages, screenshots, or code to external scanners. ClawHub listings, Snyk reports, Socket.dev analysis, VirusTotal results, npm metadata, GitHub badges, and similar signals can help a human review a mod, but they do not override Moltbank's manifest, signature, tier, permission, revocation, and user-approval gates. See [SKILL.md](SKILL.md) for the full operating rules. ## Local Development For local APP development, point `MOLTBANK_SKILL_LOCAL_PATH` at this repo (default `../moltbank-skill`) so `/skill.md` resolves to this tracked `SKILL.md` on `https://app.moltbank.bot`. Render branch-specific docs from templates: ```bash pnpm docs:render:main pnpm docs:render:preview pnpm docs:render:local ``` Validate all managed renders: ```bash pnpm docs:validate-all ``` ## Environment Variables | Variable | Purpose | Default | | :--- | :--- | :--- | | `MOLTBANK_CUSTOM_API_URL` | Explicit preview/dev API URL override. Non-production profiles live under `~/.moltbank-test`. | unset | | `MOLTBANK_CREDENTIALS_PATH` | Credentials file for the current session. | required | | `MOLTBANK_ENFORCE_DISCOVERY` | Optional schema discovery enforcement override. | CLI default | ## Requirements - Node.js >=22.0.0 - `moltbank` CLI on PATH ## Links - [Moltbank App](https://app.moltbank.bot) - [Moltbank](https://moltbank.bot) - [ClawHub](https://clawhub.ai) ## License MIT-0

moltbank

Installation

SKILL.md