task-launchlisted

Launches fresh coding sessions for in-progress tasks — one task = one session, by default a new iTerm2 tab running Claude Code; the launcher and coding tool are configurable — in the right folder, seeded with the task's context + related call transcripts (loaded fresh via the configured transcript loader inside that session) + any pass-through material + a rule to report back to the tracker when done. Use when the user wants to "launch a task", "open a terminal for this task", "spin up a task terminal", "start an in-progress task in its own terminal", "launch all my in-progress tasks", or invokes /task-launch:run — `--all` launches every in-progress task (parallel per-task analysis, ONE consolidated confirmation, one terminal each), `--one` a single task, and the default is configurable (also /task-launch:onboard to set up, /task-launch:status to health-check). Runs standalone or right after morning-brief; tracker-agnostic; config-driven via ~/.claude/task-launch/config.json. One task = one session; always a
SashaMarchuk/claude-plugins · ★ 0 · Data & Documents · score 73

Install: claude install-skill SashaMarchuk/claude-plugins

# /task-launch — one task, one fresh session You open a new session — by default an iTerm2 tab running a **fresh** Claude Code session, but the user configures both the launcher and the coding tool — for **each** task you launch (`--one`: exactly one task; `--all`: every in-progress task, one session each), in the right folder, seeded with everything that task needs: its tracker link + description, the related calls to load fresh via the configured `transcript_loader`, any Slack context worth reading, any material the user pastes (passed through **verbatim**), any follow-up questions the child must ask first, and a **report-back rule**. **Your only job is to open those sessions correctly** (right folder + right payload). What happens inside afterward is out of scope — you do not monitor them and you do not write the reports yourself; the report-back is a *rule baked into the starter prompt* that each child session follows. > Any person / project / list / folder names below are illustrative placeholders. Real values come from the user's `~/.claude/task-launch/config.json` and the shared roster `~/.claude/shared/identity.json`. This skill ships with none baked in. ## Hard rules 1. **One task = one session.** One tab / one window / one launch per task — never more. `--all` does not bend this: N tasks = N separate sessions, launched one after another. 2. **Launcher opens; it does not execute or monitor.** The report-back + status change is a rule in the child's prompt — appro