continuous-work

What problem does it solve?

Continuous-work prevents premature stopping by enforcing a “stop hook” decision gate that triggers a full, structured continuation workflow when the AI or configuration requests stopping before the job is verified and safely recorded.

Core Features & Use Cases

• Top-down “大抬头” context rebuild: performs a one-time comprehensive scan (PROFILE → requirements → Roadmap → architecture → tasks index) to ensure the next step matches the current project truth.
• Strict 9 hard-constraints execution: enforces code quality checks, engineering quality consideration, progressive disclosure of plans, deep digging when blocked, evidence-based verification when uncertain, and anti-stall behavior when blocked.
• “看得更多 / 做得更多” continuation selection: expands inspection across inward, outward, visual, deep and meta layers, then chooses the safest, most valuable next action.
• Stop-hook driven workflow across Codex/Claude/OpenCode: uses a shared decision script plus an OpenCode plugin to continue only when .ai/continuous.yaml enables it, with repeat limits and session-scoped state.

Quick Start

Ask your AI to follow the continuous-work execution rules before stopping, with your project’s stop configuration enabled via .ai/continuous.yaml.