hs:afk

What problem does it solve? Executing an approved plan or PRD autonomously is risky: unattended agents can push broken code, bypass review, or run unsandboxed on the host machine. This Skill provides a guarded AFK workflow that preflights readiness, routes execution into a Docker-sandboxed Ralph loop, and keeps humans reviewing at both ends. ## Core Features & Use Cases - Preflight readiness check: Runs a fail-open preflight script that reports ok/warn/fail findings with fix hints before any autonomous work begins. - Sandboxed autonomous loop: Routes green runs to a Ralph Docker sandbox where the loop only commits locally — never pushes, ships, or deploys. - Native fallback path: When preflight fails, offers a plain-language 3-option choice including a host-based fallback with an explicit warning banner about lost isolation. - Use Case: You have a human-approved implementation plan and PRD. Invoke the skill with the plan, PRD, and a small iteration count; it verifies Docker readiness, runs the sandboxed loop committing per iteration, and stops for your diff review before anything ships. ## Quick Start Ask the AI to run the approved plan and PRD in AFK mode with a small number of loop iterations, for example by invoking /hs:afk with the plan file, PRD file, and N set to 3.