sancai-tian

Seed ideas, render panoramas, and compress graph state in persistent JSON.

Updated Mar 29, 2026
One-click install
npx skills add https://github.com/songlairui/agent-monad --skill sancai-tian
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sancai-tian
Source: https://github.com/songlairui/agent-monad/tree/main/skills/sancai-zhen/tian
Command: npx skills add https://github.com/songlairui/agent-monad --skill sancai-tian

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill keeps AI workflow context organized when ideas, actions, and questions accumulate across sessions, so the graph stays understandable instead of fragmenting into scattered notes.

Core Features & Use Cases

  • Seeding: Add a new idea or seed, infer whether it belongs to a project or matter, and write it into the persistent graph.
  • Panorama Rendering: Load the graph and show a clear overview of projects, matters, open nodes, and the most important emerging signal.
  • Compression: Merge duplicates, archive resolved items, and summarize long-running fragments when the graph grows too large.
  • Use Case: A multi-agent coding session leaves behind progress, questions, and insights; this Skill turns that history into a shared, agent-agnostic context layer.

Quick Start

Ask the AI to automatically seed the current idea into the harness graph and then render the panorama if you want a full overview.

Frequently Asked Questions about sancai-tian

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I keep AI context organized across multiple sessions?

Keeping AI context organized across multiple sessions requires a persistent JSON graph that seeds ideas, renders the current panorama, and compresses accumulated state. This prevents projects and matters from fragmenting into scattered notes during agent handoffs.

What is context compression in a multi-agent AI workflow?

Context compression in a multi-agent AI workflow is the process of merging duplicate nodes, archiving resolved items, and summarizing long-running fragments. It uses confidence-based routing and confirmation-gated batch updates to keep the shared graph understandable.

How do I seed new ideas into a persistent workflow graph?

To seed new ideas into a persistent workflow graph, you add an idea or seed and let the system infer whether it belongs to a project or matter. It then writes the data into the persistent JSON graph for later panorama rendering.

Does multi-session AI workflow context require manual graph loading and saving?

Yes, multi-session AI workflow context requires persistent JSON graph loading and saving. The system relies on this manual or automated persistence to maintain a shared, agent-agnostic context layer across discovery and cleanup phases.

What is the best way to manage an accumulating context graph in AI projects?

The best way to manage an accumulating context graph in AI projects is to apply confirmation-gated batch updates for compression and archival changes. This merges duplicates and archives resolved items when the graph grows too large.

When should I not use a compression-based context graph for AI handoffs?

You should not use a compression-based context graph for AI handoffs if your work lacks persistent JSON graph loading and saving capabilities. The system requires confidence-based routing and batch updates to safely merge duplicates and archive resolved items.