guanbearguanbearCommunity·1 Agent Skills Included

OctoClaw

Automatic task delegation, status tracking, and cost-aware model routing

Delegates slow or heavy tasks to background sub-agents so the main agent stays responsive. Routes each request through a cheap local judge model to cut expensive frontier-model token costs. Keeps the main conversation clean by returning only compact result packets instead of raw transcripts. Reports trustworthy task status and delivers results to Slack, Feishu, WeChat, and the CLI.
npx skills add guanbear/OctoClaw --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install OctoClaw?

Run `npx skills add guanbear/OctoClaw --all -g -y` in your terminal to install the skill globally. You need OpenClaw 2026.5.12 or newer and Node.js 22 or newer.

How to stop my main agent from stalling on big tasks?

OctoClaw automatically decides per turn whether to reply directly or delegate the work to a background sub-agent, so the main agent stays responsive and you always see live status.

How to reduce AI model costs on simple requests?

A lightweight local judge model (like Qwen3 0.6B via Ollama) routes each request, so your expensive frontier model is never used just to decide how to handle a simple ask.

Does OctoClaw work with Slack and Feishu?

Yes. It delivers task acknowledgments, progress updates, and final results to Slack with streaming, Feishu threads, and WeChat plain text, plus a full CLI for status checks.

Can I track whether a delegated task actually finished?

Yes. Task lifecycle truth comes from OpenClaw's native TaskFlow with a SQLite audit ledger, so states like running, stalled, timed out, or completed always have verifiable evidence.

Related Repositories in Software Engineering

View All in Software Engineering