← ClaudeAtlas

openrouterlisted

Call other models through OpenRouter with one small script. Use for web research with citations, current facts, a second opinion from another model, embeddings, a closed-choice decision, "which model should answer this", model prices or context sizes, or credit left on the key. When no model is named, Jev picks one per request, and every call reports the model that answered and what it cost.
opascope/skills · ★ 0 · AI & Automation · score 75
Install: claude install-skill opascope/skills
# Ask another model through OpenRouter Read [shared.md](shared.md) first. Save each call's receipt with kind `receipt`. The script sits beside this file. Run it by its full path: `python3 "<this skill's directory>/scripts/openrouter.py" <verb> ...`. Resolve the directory from where you loaded this SKILL.md; do not rely on a variable for it. ## When to use it, and when not Use it when the answer needs something you do not have: the live web, a different model's view, a vector, a decision limited to fixed options, or catalog facts. Do not use it for ordinary reasoning or writing you can do yourself. A second model is a cost, not a ritual. ## The key Networked verbs need `OPENROUTER_API_KEY` in the environment. Create one at https://openrouter.ai/keys. `models` works without it. Never print the key, put it in a command line you show, or save it in any file. ## The verbs ``` openrouter.py chat "<prompt>" [--role R | --model ID] [--system S] [--json] [--max-tokens N] [--effort none|low|medium|high] [--image PATH] openrouter.py research "<question>" [--deep] openrouter.py pick "<representative prompt>" openrouter.py decide --state FILE|- --questions FILE openrouter.py embed "<text>" [--json] openrouter.py models [list [--search Q] [--newest] | check] openrouter.py key ``` Every networked verb takes `--json-out` for one JSON object, and `--receipt FILE` to write the trailer and the generation id to a file. ## How the model is chosen First mat