free-scout

Runs read-only scouting prompts on free OpenRouter models via headless Claude CLI subprocesses.

Updated Sep 8, 2026
One-click install
npx skills add https://github.com/salomepoulain/makery-stations --skill free-scout-salomepoulain
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: free-scout
Source: https://github.com/salomepoulain/makery-stations/tree/main/stations/claude/workbench/pantry/skills/free-scout
Command: npx skills add https://github.com/salomepoulain/makery-stations --skill free-scout-salomepoulain

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Spawning a full Sonnet subagent for every small lookup burns through Claude Pro usage-limit quota. This Skill offloads small, well-scoped, read-only tasks to free OpenRouter models running in a separate headless process, preserving quota for work that actually needs a top-tier model. ## Core Features & Use Cases - Free-model scouting: Runs a single self-contained prompt on free OpenRouter models (deepseek-r1:free, qwen-qwq:free) through a headless claude -p subprocess instead of a Sonnet subagent. - Fixed read-only sandbox: Grants only Read, Grep, Glob, and safe non-mutating Bash commands, with Write/Edit and mutating commands explicitly denied. - Model rotation with fallback: Rotates across configured models on rate-limit or empty responses, and returns distinct exit codes so the caller falls back to a real subagent when OpenRouter is unavailable. - Use Case: You need to locate where a function is defined or check whether a pattern exists in the repo. Instead of spawning an Explore subagent, run free-scout in the background with a self-contained prompt and continue other work while it answers. ## Quick Start Run the free_scout.py script in the background with a fully self-contained read-only question, then check its exit code and fall back to a normal subagent if it is non-zero.

Frequently Asked Questions about free-scout

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a read-only repo query on a free OpenRouter model?

Run scripts/free_scout.py with a fully self-contained prompt as its argument, preferably as a background task. The script spawns a headless claude -p subprocess pointed at OpenRouter and prints the answer to stdout on success.

When should I use a free model instead of a Sonnet subagent?

Use it for small, read-only, low-judgment tasks like locating code, summarizing a file, or checking whether a pattern exists. Avoid it for multi-step reasoning, anything requiring writes, or tasks where a wrong answer is costly.

What happens if OPENROUTER_API_KEY is not set?

The script exits with code 2 and prints a fallback message to stderr. The caller must then run the task with a normal Explore or general-purpose subagent and inform the user that quota was not saved.

Can the scout process edit or write files?

No. The temporary settings grant a fixed read-only baseline of Read, Grep, Glob, and safe non-mutating Bash commands, while Write, Edit, NotebookEdit, and mutating commands like rm or git push are explicitly denied.

Which free OpenRouter models does the script use?

The config.json file lists deepseek/deepseek-r1:free and qwen/qwen-qwq:free. The script rotates to the next model with a short backoff when one hits a rate limit or returns an empty or malformed response.