Kiro Ecosystem Graph

See your AI's brain. Fix what's broken.

Interactive force-directed graph that visualizes your Kiro cognitive ecosystem — steerings, skills, hooks, and their interconnections rendered as a neural network brain map.

Author: Weberson Rodrigues

Quick Start

1. Install the extension
2. Open a workspace with .kiro/steering/ files
3. Click the brain icon in the activity bar

Or use Command Palette: Ecosystem Graph: Show

📖 Building Your Cognitive Ecosystem

New to AI cognitive ecosystems? Read the complete step-by-step guide:

→ How to Build an AI Cognitive Ecosystem for Any Project

Learn the 9 layers, the two fundamental pillars (Instruction + Access), hook automation patterns, and how to evolve your system from zero to full autonomy.

Why This Exists

Your AI agent is only as good as its knowledge network. Without visibility:

• Steerings exist but the agent doesn't find them at the right moment
• The same rule is documented in 3 places, sometimes contradicting itself
• Files created months ago sit unused — dead knowledge rotting silently
• A single backtick reference is the only link between two critical documents

This extension makes the invisible visible. Every orphan node is knowledge your agent can't reach. Every hub with 30+ connections is the backbone of your system's intelligence.

Real Results

A team with 37 steering files achieved after using the graph to identify and fix gaps:

• 44 nodes, 133 edges, 10 types — densely connected knowledge network
• 0 orphan steerings — every piece of knowledge is reachable
• Hub node with 38 connections — central design system doc connects to everything
• 3 fragile links identified and fixed — single-reference connections strengthened
• 4 workspace folders covered — API, Mobile, Integrator, Infrastructure

The graph revealed their authentication flow was connected to only 1 other document. After adding cross-references, the agent went from "I think the auth uses..." to precise, documented answers — every time, instantly.

Cognitive Analysis

Click the brain button in the toolbar to open the analysis panel:

• Orphan Steerings — files with zero connections (isolated knowledge)
• Fragile Links — single-reference connections that could easily break
• Isolated Files — nodes with no references to or from anything
• Coverage Gaps — workspace folders with no steering files
• Suggestions — actionable recommendations to strengthen the network

Click any node name to center and highlight it in the graph.

Visual Features

The Difference

All Features

• Force-directed graph with neural network aesthetic
• Multi-file discovery — steerings, skills, hooks (.json and .kiro.hook)
• Reference extraction — wiki-links, markdown links, backtick refs, table paths, hook prompts
• Brain Health Panel — orphan count, hub detection, cognitive weight, islands
• Filter panel — toggle by node type and workspace folder
• Real-time updates — file watcher with 500ms debounce
• Click-to-open — click any node to open the file
• Edge hover tooltip — reference count and types
• Enhanced hook tooltips — trigger type, description, referenced steerings
• Settings panel — real-time physics controls (forces, sizes, labels)
• Multi-workspace support
• Fully offline — no network, no telemetry

Node Types

Requirements

• Kiro or VS Code 1.85+
• Workspace with at least one .kiro/steering/*.md file
• No external dependencies — fully offline

Settings

Adjustable in real-time via the gear icon:

• Physics: center force, repulsion, link strength, link distance
• Display: node size, link width, arrows, label mode
• Filters: show only existing files, show orphan nodes

License

MIT — Free forever.