Skip to content
Conduit

skill

code-explorer

A skill for Claude Code that deep-analyzes any source code repository and produces a single self-contained interactive HTML dashboard — knowledge graph, architecture overview, data/logic flow, depende

raghavishwas/code-explorercommunitylistedunreviewed
Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/raghavishwas/code-explorer itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 27, 2026

Code Explorer — Parallel Multi-Agent Edition

A skill for Claude Code that deep-analyzes any source code repository and produces a single self-contained interactive HTML dashboard — knowledge graph, architecture overview, data/logic flow, dependency map, key functions & classes, and tech stack breakdown.

It fans out 5 specialist subagents in parallel, each analyzing one dimension of the repo independently, then synthesizes everything into one dashboard. Target wall time: ~2–3 min regardless of repo size.

What it produces

A single code-explorer-dashboard.html with 6 tabs wired to real analysis data:

TabKey feature
Knowledge GraphInteractive D3 force-directed graph
ArchitectureLayered diagram + per-layer cards
Data FlowSwimlane step-by-step flows
DependenciesBar chart + sortable table
SymbolsSearchable function/class table
Tech StackCard grid by category

All data is embedded as JS constants — no network calls at runtime.

How it works

┌─────────────────────────── LEAD AGENT ───────────────────────────┐
│  Phase 1: Quick orient → write shared manifest                    │
│  Phase 2: Fan out 5 subagents simultaneously (background)         │
│  Phase 3: Collect JSON results from all agents                    │
│  Phase 4: Build dashboard → present                               │
└───────────────────────────────────────────────────────────────────┘
     [agent-1]   [agent-2]   [agent-3]   [agent-4]   [agent-5]
     Tech Stack  Knowledge   Data Flow   Dep Map    Key Symbols
     & Arch      Graph       & Logic     Analysis   & Patterns

The five subagents are read-only — they never edit your source files.

Installation

Clone into your Claude Code skills directory:

https://github.com/raghavishwas/code-explorer.git \
  ~/.claude/skills/code-explorer

Or for a single project, place it under .claude/skills/ in the project root.

Usage

Inside a project folder, ask Claude Code any of:

  • "explore this repo"
  • "analyze my codebase"
  • "knowledge graph of my code"
  • "explain this project"
  • "map my source code"
  • "visualize my repo"

The skill auto-triggers and delivers the full dashboard.

Repo layout

SKILL.md                       # Skill manifest + orchestration flow
agents/
  tech-stack-arch.md           # Agent 1 — tech stack & architecture
  knowledge-graph.md           # Agent 2 — entity/relationship graph
  data-flow.md                 # Agent 3 — data & logic flows
  dep-map.md                   # Agent 4 — dependency map
  key-symbols.md               # Agent 5 — key functions/classes
references/
  dashboard-spec.md            # Dashboard visual spec + D3 implementation

Tips for large repos (>500 files)

  • Each agent focuses on its domain's most important 30–50 files.
  • Set CLAUDE_CODE_FORK_SUBAGENT=1 for ~90% token cost reduction on child agents.
  • Knowledge graph targets 30–70 nodes; leaf-only utility files are pruned.

License

MIT

Disclaimer

Not affiliated with or endorsed by Anthropic. "Claude" and "Claude Code" are trademarks of Anthropic, referenced here only to describe compatibility.

↑↓ navigate openesc close