What it is
A single self-contained HTML page that visualizes my entire Claude Code configuration: 11 custom subagents, 30+ skills and commands, PostToolUse hooks, plugins, and the CLAUDE.md rules that orchestrate them. The dashboard is generated and incrementally updated by a custom /dashboard skill — Claude scans its own configuration files, diffs them against the page, and edits only the sections that changed.
Highlights
- Self-maintaining — an incremental-update skill compares badge counts against the real config and touches only changed sections, appending a changelog entry per run
- Agent orchestration map — 11 role-specialized subagents (developer, debugger, architect, verifiers, …) with a color-coded delegation scheme
- GPT cross-validation loop — plans and core code changes are adversarially reviewed by a second LLM (GPT via Codex CLI) with a file-based audit trail, before implementation proceeds
- Hook automation — PostToolUse hooks enforce writing style and auto-trigger cross-validation when a plan is finalized
- Sanitized publishing pipeline — a script strips private data (research memory, workspaces, local paths) from the canonical dashboard to produce this public copy
Why
Working with AI agents is only productive when the configuration is legible. This page is the answer to “what exactly is my setup right now?” — kept honest because the same AI that uses the configuration is the one that documents it.