← ClaudeAtlas

taskslisted

How to use taskmgr — the file-based task tracker: its data model, core commands, and tracking discipline.
hk9890/claude-dev-tools · ★ 0 · AI & Automation · score 73
Install: claude install-skill hk9890/claude-dev-tools
# Using taskmgr `taskmgr` is a lean, file-based task tracker: issues, dependencies, and ready-work as Markdown files in a store. The common shape is a `.tasks/` directory versioned alongside the code; a store can also be **central** (`taskmgr init --central`, or a named store), living outside the project — `taskmgr where` tells you which one you are on. You operate it **only** through the `taskmgr` CLI — the `.tasks/` directory is an internal store, never a place to read or edit by hand. Do **not** `cat`, `grep`, `ls`, or write the files under `.tasks/` directly: the CLI holds the write lock, resolves the store by walking up from any subdirectory, and emits stable `--json`. Touching the files directly bypasses the lock (risking corruption), misses stores that live above the current directory, and couples you to a format that can drift. This skill covers the model and the discipline; for the exact flag surface, ask the tool itself (see "Source of truth" below). ## 1. Is taskmgr available? Before using the tracker, confirm the binary and a store exist — probe them **separately** so the two failure modes stay distinct: ```bash command -v taskmgr >/dev/null 2>&1 # 1) is the binary installed? taskmgr where # 2) does a store resolve, and which one? ``` - **`command -v taskmgr` fails** → no binary. Stop and tell the user to install it (`make install` from the task-manager repo). Do not fall back to TodoWrite or markdown files for tracking. - **`taskm