Grok Research (OpenRouter + Grok 4.1)

This skill automates a source-backed “deep dive” on a recent topic using Grok 4.1 via OpenRouter. Output is a Markdown report with citations.

Quick start

1. Export the API key:

export OPENROUTER_API_KEY="..."

2. Run the script:

node skills/grok-research/scripts/grok-research.js --topic "What's going on with …" --out report.md

By default it tries to use x-ai/grok-4.1-fast:online. If your account/model uses a different id, pass --model or set GROK_RESEARCH_MODEL.

Environment variables

• OPENROUTER_API_KEY (required)
• GROK_RESEARCH_MODEL (optional) (default: x-ai/grok-4.1-fast:online)
• OPENROUTER_SITE_URL (optional) (recommended header by OpenRouter)
• OPENROUTER_APP_NAME (optional) (recommended header by OpenRouter)

Useful flags (script)

• --topic "...": topic to research (or first positional argument)
• --out report.md: saves the report and prints the resulting path to stdout
• --model <id>: forces the OpenRouter model id
• --max-tokens 2200, --temperature 0.2: length/style controls

Recommended workflow (agent)

1. Clarify the goal: “do you want an executive summary or a decision-making report?”
2. Ask for constraints: country/language, time window (last 24/72h), and any specific claims to verify.
3. Run scripts/grok-research.js and review the Claims & verification section:
   • If sources are weak/missing, re-run with higher --max-tokens or narrow the topic.
4. If the topic is sensitive/controversial, ask the model to list alternative hypotheses and what evidence would falsify/confirm each.

Resources

• Main script: scripts/grok-research.js
• API notes: references/api_reference.md