← ClaudeAtlas

tasks-worklisted

Run ready taskmgr tasks end to end — implement → verify → record, then verify (never auto-close) the parent epic.
hk9890/claude-dev-tools · ★ 0 · AI & Automation · score 76
Install: claude install-skill hk9890/claude-dev-tools
# Running ready work Take ready taskmgr tasks through implementation, verification, and recording — driven by the bundled `work.js` workflow. You confirm the scope here in the main loop, then hand the resolved task ids to the workflow, which runs one `implementer` per task sequentially, verifies each (review ∥ test), and records the outcome. Epics are verified and left for a human to close. **Scope (optional):** an epic id or task ids passed with the command pre-seed the step 3 selection — still confirmed there before anything runs. $ARGUMENTS ## 1. Preconditions First, **load the `tasks` skill** for the CLI surface and the taskmgr gotchas it relies on (closure is not gated, concurrent writes are safe). Then confirm the tracker is usable: run the availability check from that skill ("Is taskmgr available?") — if it fails, stop and tell the user. ## 2. Discover ready work ```bash taskmgr ready --json # open tasks with no open blockers — priority, then oldest ``` If nothing is ready, report that and stop. Otherwise present the ready queue briefly (id, type, priority, title). ## 3. Confirm scope Ask the user via `AskUserQuestion` which scope to run — when the command carried an argument, present that scope as the recommended option instead of an open question: - **All ready** — every task `taskmgr ready` returned. - **A subset** — specific ids the user picks. - **One epic** — run that epic's ready children, then verify the epic. Resolve the selection to a concr