Word Aligner agent skill
Install the skill once, then ask your assistant in plain language. It translates, works out which words correspond, calls the Word Aligner API, and hands back a shareable diagram link. No API key, no sign-up.
What it is
The skill is a small package that teaches an AI assistant how to use Word Aligner. It holds a SKILL.md with instructions and worked patterns, an API reference, and an OpenAPI schema. Any assistant that can follow instructions and make an HTTP request can run it, so the same package works across Claude, ChatGPT, and custom agents.
What you can ask
Once it is installed, requests like these produce a diagram link:
- “Translate ‘я хочу спать’ into English and show the word alignment.”
- “Make an interlinear gloss of this Latin sentence.”
- “Align this Hebrew verse with its English translation, right to left.”
- “Show which French words map to which English words in this pair.”
The assistant returns a Word Aligner URL you can open, edit, or export as PNG, SVG, or PDF.
Install
Claude
Add it as an Agent Skill. In Claude Code, put the word-aligner-skill/ folder in your skills directory (~/.claude/skills/ for personal use, or .claude/skills/ in a project). In the Claude apps that support skills, upload word-aligner-skill.zip where custom skills are added. As a fallback on any Claude surface, paste the contents of SKILL.md into your project instructions.
ChatGPT
Build a Custom GPT. Add an Action from the OpenAPI schema at /api/align/openapi.json, then paste SKILL.md into the instructions so the model knows how to count word indices and lay out a gloss. No authentication is needed.
Any other agent
The API is open and needs no key, so any tool-using agent can call it. Point your framework at POST /api/align with the SKILL.md guidance and the full API reference. The package also ships an agents/openai.yaml manifest for OpenAI-style agents.
How it works
Under the hood the skill calls one endpoint, POST /api/align, with the text lines and the word pairs that link them, and gets back a URL:
{
"lines": ["Hello world", "Bonjour le monde"],
"alignments": [[0, 0, 1, 0], [0, 1, 1, 1], [0, 1, 1, 2]]
} The hard part for a model is counting word indices and laying out a gloss tier, which is what SKILL.md spells out. The API reference covers every parameter: per-line fonts and RTL, palettes and line style, and pair-level gap and connector control.