A local work board and orchestration loop for Claude Code. Every task saved as a ticket survives context compaction and new sessions, with instructions, status, and recorded discussion kept on the board. Categories route tickets to a model and effort; token-gated executors return verified work for review and integration. A self-hosted Kanban dashboard spans your projects.
5 skills · 19 agents · 1 MCP server · hooks
Set up a project from your existing Claude Code habits, keep Toolshed installs healthy, and find missing capabilities after real work accumulates. Local scripts reduce session history to a bounded summary before the active model sees it; setup, resupply, updates, and config changes happen only with your approval.
4 skills · hooks
Your ChatGPT/Codex and Grok subscription models in Claude Code's /model picker. A local gateway routes claude-gpt-* models to OpenAI's Codex backend, claude-grok-* models to Grok, and everything else to api.anthropic.com, so you switch between Claude, GPT-5.x, and Grok mid-session like any other model. No API keys.
3 skills · hooks
A self-maintaining codebase map: atomic docs Claude loads at the start of each session and keeps in sync as the code changes. Any stack; greenfield or existing.
2 skills · hooks
Atomic project rules in .claude/live-rules/rules/ with a generated manifest. Claude injects applicable rules at SessionStart, then only new or changed unseen rules as prompts and edits match; legacy single-file storage is for migration or explicit overrides.
2 skills · hooks
Local usage board for Claude Code. A hooks module records one metadata row per turn (exact engine cost, tokens by model, tool calls, agent spawns, context fill, rate-limit windows) into a per-project JSONL ledger under your Claude config directory, and /observability:board opens a self-contained HTML board. Recording is off until /observability:track on in a project. No server, no network, nothing leaves the machine.
1 skill
Local JavaScript and TypeScript function complexity and measured V8 coverage, with per-function CRAP scores strictly below 6. The scorer reads source with the TypeScript compiler installed in the scored project (or its own dependency), pairs each function with V8 coverage by name or source position, and prints a per-function report. Everything runs on your machine; no source, coverage or report leaves it.
1 skill