qq-go-afk-ham

The unbounded end of the spectrum, quality-first, budget isn’t the constraint.

The problem: A big, ambitious, multi-day build either needs you at the keyboard the whole time, or it runs unattended and you come back to a session that died five hours in, a queue nobody touched since, and no reliable way to tell what got built versus what an agent merely claimed.

This is the unbounded, max-effort half of the AFK pair (sibling: qq-go-afk-smart): resilience stood up first so the campaign survives caps and session death and resumes in minutes, real strategic thinking instead of a flat task dump, workflows spawning workflows for the execution itself, and an audit gate on every key step so nothing ships on a self-report. Reach for it when the work is open-ended and exploratory and the value is in finding a compounding advantage, not grinding a fixed list.

For the frugal end, use qq-go-afk-lean. For the adaptive middle, use qq-go-afk-smart.

Key principles

• Resilience first, then strategy, then surfaces, then execution: the order isn’t negotiable
• Audit gate at every key step, no exceptions; nothing ships with a real issue still open
• A server-side rate limit is a retry, not a wall; back off, then resume only the failed units, never the whole run
• Verify the felt outcome, not the inventory; a status API reporting “healthy” and the real data path working are two different questions
• Count only what’s verified on disk, never a number an agent merely claims

Part of Operator Skills, a collection of skills for running a real Claude Code setup.