Build a confidence-scored execution plan from ambiguous user intent and project context.
npx skills add https://github.com/songlairui/agent-monad --skill sancai-di
Please help me install this Agent Skill. Skill: sancai-di Source: https://github.com/songlairui/agent-monad/tree/main/skills/sancai-zhen/di Command: npx skills add https://github.com/songlairui/agent-monad --skill sancai-di
This Skill turns vague or complex work requests into a structured consensus page so the AI can decide what it can execute immediately, what needs clarification, and what must be confirmed.
Use the sancai-di skill to evaluate this request, build the consensus page, and carry out the task with confidence-based autonomy.
High-intent search queries and answers about installing and using this skill.
This Skill turns vague work requests into a structured consensus page by capturing intent, current state, boundary, and risk, enabling AI to decide what to execute immediately, clarify, or confirm during multi-step development workflows.
Confidence scoring applies thresholds to intent, done-state, current-state, boundary, and risk assessments, routing work into autonomous execution when confidence is high and triggering targeted follow-up questions when it is not.
Yes, this Skill supports bug fixes and refactors by building a confidence-scored consensus page that defines target state, current state, boundary, and risk, enabling adaptive questioning and autonomous progress reporting.
A consensus page in autonomous agent workflows is a structured artifact capturing intent, target state, current state, boundary, and risk, used to drive high-confidence autonomous execution and manage task handoffs safely.
You should use confidence thresholds for safe interruption when an agent workflow encounters ambiguous user intent or elevated risk, requiring confirmation or clarification before proceeding with autonomous execution.
Yes, this approach works without external dependencies for software change requests. It internally evaluates intent and project context to build a consensus page, managing execution, clarification, and confirmation autonomously.
cshyangCreate structured phase checkpoints documenting state, completed work, and upcoming tasks.
liauw-mediaDispatch fresh subagents sequentially for plan-driven tasks with mandatory reviews.
FVossebeldCoordinate multi-agent systems with supervisor routing, handoffs, and failure handling.
sburlDelegate coding tasks to the Gemini CLI for autonomous execution.
