← ClaudeAtlas

next-cardlisted

Pick the highest-leverage open card to work on next. Read-only verdict — does NOT flip status. AUTO-INVOKE when user says "what's next", "pick something", "work on the queue", "what should I do", "next item", "drain the deck", or initiates autonomous-loop work. Filters to `human_gate: none` for loop safety; Kanban pull principle (work is taken, not pushed).
zauberzeug/game-of-cards · ★ 2 · AI & Automation · score 68
Install: claude install-skill zauberzeug/game-of-cards
## Preflight If any `!` block below shows `goc: command not found`, `Permission for this action has been denied`, or `no such file or directory: .game-of-cards/deck/`, **stop and invoke `Skill(kickoff)` first**. Kickoff detects which setup step is missing (CLI not installed, Bash allowance not granted, project state not scaffolded) and walks the user through it. Re-invoke this skill only after kickoff completes. ## Context !`goc --status active -v 2>&1 | head -20` !`goc --ready -v${GOC_WORKER:+ --worker "$GOC_WORKER"}` # Pick the next card Kanban's **pull principle** (Anderson): work is taken when capacity is ready, not pushed by a planner. The runner — human or autonomous loop — pulls one card at a time, top of the value-sorted queue, gated by the autonomy ladder so a cron run cannot accidentally land a research-impacting decision unsupervised. The deck is the queue; this skill is the picker. Recommend, do not claim. **Status does NOT flip here** — that is `Skill(advance-card)`'s contract. The pull is a two-step gesture (pick, then claim) so the human or /loop can abort between them without half-state on disk. Before recommending, read the active slice above. Active cards are claimed soft locks; avoid recommending the same card or adjacent/conflicting work unless the user explicitly asks to continue that active card. User argument: $ARGUMENTS — if non-empty (a title or area like an area tag or path prefix), narrow the queue. If empty, scan the full ready slice (`--r