← ClaudeAtlas

tipatask-expertlisted

Tipatask project expert. Guides tag application to tasks, architecture file navigation, and system-specific tag management. Always active for this project. Load when: creating tasks, working on a task, reading architecture, applying tags, looking up system modules, asking about project structure.
Tipatask/tipatask · ★ 1 · Data & Documents · score 72
Install: claude install-skill Tipatask/tipatask
# Tipatask Expert System-specific tags (`tt-*`) are the primary navigation mechanism for the architecture knowledge base at `ai/architecture/`. Every task should carry the right `tt-*` tags. Every implementation session should load the matching architecture files before touching code. ## MCP Server Split Two MCP servers are registered — call each tool on the right one: - **`tipatask`** (remote HTTP, primary) — architecture and tag tools: `list_system_tags`, `get_tag_architecture`, `get_tag_architectures`, `get_project_tags`, `create_system_tag`, `ensure_project_tag`, plus task CRUD except guarded completion. - **`tipatask-local`** (stdio, this machine only) — 5 tools that need a repo checkout, local KB files, or Git state the remote API doesn't have: `batch_grep_tags`, `push_knowledge`, `pull_knowledge`, `git_worktree_status`, `complete_task`. Use `complete_task` for guarded completion. If a tool call fails with "tool not found," check you called it on the right server name. ## Credential recovery Project config holds only the API target; the account store under `TIPATASK_USER_DATA` holds credentials. Follow `CLAUDE.md` § Credential recovery and completion, including `tt verify` / `tt complete` through the shipped credential-safe CLI when remote MCP is unavailable. Missing user-data paths are not expiry. Never print tokens or bypass local completion verification with a REST status write. ## Architecture Context Loading **Before implementing anything:** 1. Read `ai/ar