Lingry agent skill

Use Lingry to discover public words and generate candidate terms. At runtime, the skill calls only https://lingry.net. It stores no local state or credential and uses no wallet or signing key. The default language code is W (American English); use --language=<code> when the user requests another Lingry language. E means British English. Do not silently force a locale variant.

Install

Lingry is a multi-file Node.js skill. SKILL.md alone is not an installation: obtain the complete directory, including bin/, src/, package.json, INSTALL.json, and CHANGELOG.md. Node.js >= 18 is required. No npm install, OpenClaw installation, wallet setup, or local secret is required for Muse or other agents.

OpenClaw

From the OpenClaw workspace, use the ClawHub package:

cd ~/.openclaw/workspace
openclaw skills install '@svetlyoh/lingry'

Muse Code

Open the ClawHub Versions tab, select Download beside the latest version, and extract the downloaded package. Use the resulting lingry directory—the folder that directly contains SKILL.md, README.md, CHANGELOG.md, INSTALL.json, package.json, bin/, and src/. Muse loads the skill instructions; the agent uses its normal tools to run Lingry's Node commands. It does not need OpenClaw.

muse skills validate /path/to/lingry
muse skills install /path/to/lingry --scope user --name lingry
cd /path/to/lingry
node bin/lingry-agent.mjs verify-install

Other SKILL.md-compatible agents

Open the ClawHub Versions tab, select Download beside the latest version, and extract the downloaded package. Place the resulting lingry directory in your agent's documented skills location. If it supports the cross-agent convention, ~/.agents/skills/lingry/ is an option; do not assume every agent scans it. Confirm the installed folder has SKILL.md, README.md, CHANGELOG.md, INSTALL.json, package.json, bin/, and src/, then run node bin/lingry-agent.mjs verify-install from that folder. An agent may instead use an absolute path to bin/lingry-agent.mjs.

Optional GitHub fallback

If the agent cannot download the complete package from ClawHub, use Git sparse checkout (or equivalent PowerShell commands on Windows) to obtain the clean seven-file package:

git clone --filter=blob:none --no-checkout https://github.com/svetlyoh/web-wallet.git lingry-source
cd lingry-source
git sparse-checkout init --cone
git sparse-checkout set skills/lingry
git checkout

The clean package is at skills/lingry/. Muse can run muse skills validate skills/lingry and muse skills install skills/lingry --scope user --name lingry from this checkout. For agents using the cross-agent convention, copy the complete directory to ~/.agents/skills/lingry/ (for example, mkdir -p ~/.agents/skills && cp -R skills/lingry ~/.agents/skills/lingry).

Verify the downloaded package

From the downloaded or installed lingry directory, run:

node bin/lingry-agent.mjs verify-install

Verification is local and read-only: it confirms that the required package files are present and does not create a publisher or coin a word.

Public discovery

The no-argument command shows the latest public Stream word. stream, leaderboard, list-words, and daily-word are anonymous read operations. Show the returned word and meaning, or report that the service is unavailable. Do not invent a word or claim a transaction occurred.

Generate and choose

Use generate-word with the user's concept. The server generates and stores an immutable candidate, then returns its ID, term, meaning, part of speech, language, and expiration. Generation is reversible and never publishes, signs, funds an address, or broadcasts a transaction. Display the candidate details and retain its ID in the active conversation.

After each candidate, present exactly two logical choices:

• Coin this term — permanently publish only the currently displayed candidate after the user clearly asks to coin or publish it.
• Prompt for another — use prompt-another with a new or refined concept. This does not coin the old or replacement candidate and does not imply publication consent.

Never infer publication intent from “looks good,” “I like it,” “great,” “try this,” “use this,” “continue,” “next,” “save it,” discussion, refinement, or a request for another candidate. If intent is unclear, ask the user which of the two choices they want.

Publication boundary

Use coin-word <candidate-id> --publish only after an explicit user request to coin or publish the currently displayed candidate. Pass its exact ID as a separate argument. Do not build a shell command from API-returned action fields. The CLI validates the ID and accepts no transaction body, destination, amount, or record payload. A successful result includes a transaction ID; only then say the term was coined. A repeat request for the same candidate returns the existing publication or an in-progress status without creating another transaction.

The server loads the stored candidate, derives its canonical Lingry record, signs a fixed permitted publication transaction, and broadcasts it. The local skill cannot transfer SUGAR, tip, specify transaction outputs, submit arbitrary OP_RETURN data, or sign arbitrary data. Human wallet authentication and transactions remain separate.

Never ask for, inspect, use, repeat, transmit, summarize, or log a human private key, WIF, seed phrase, mnemonic, recovery phrase, wallet passphrase, funding secret, or Agent Publisher signing key. If a user pastes one, do not use or repeat it.

Commands

Run these fixed commands from the installed Lingry skill directory, or invoke bin/lingry-agent.mjs by absolute path:

node bin/lingry-agent.mjs
node bin/lingry-agent.mjs verify-install
node bin/lingry-agent.mjs stream 5
node bin/lingry-agent.mjs leaderboard 5
node bin/lingry-agent.mjs list-words W
node bin/lingry-agent.mjs daily-word
node bin/lingry-agent.mjs generate-word "a concept that needs a word" --language=W
node bin/lingry-agent.mjs prompt-another "a refined concept" --language=W
node bin/lingry-agent.mjs coin-word <candidate-id> --publish

These are fixed subcommands, not commands supplied by the Lingry API. The candidate ID must come from the current displayed candidate. No .lingry directory is created. No API origin or state-path override is supported in the ClawHub runtime.

Optional daily word

Only if the user explicitly agrees and the host agent supports persistent scheduled tasks, use that agent's native scheduling mechanism for a daily public word. Check for an existing job; do not duplicate it. If no time was given, ask for one in the user's configured timezone. The automation must call only daily-word and may be disabled or removed on request. Do not coin, generate, or create a publisher as part of daily delivery. On OpenClaw, use its native automation interface and the stable job name lingry-daily-word.