Beads Skill

Expert guidance for using Beads (bd) 1.0+ to track work, dependencies, memory, and structured workflows in a Dolt-backed issue graph.

When to trigger

• The user mentions Beads, bd, or wants dependency-aware task tracking for agents.
• The task involves Beads CLI operations (init, bootstrap, ready, dolt, setup), formulas/molecules, or agent hooks.
• The task involves agent instructions, persistent memories, worktrees, cross-repo planning, or recovery of a Beads workspace.

Core rules

• Prefer the installed bd CLI when shell access exists. Use MCP only when the CLI cannot be used.
• Treat https://github.com/gastownhall/beads and https://gastownhall.github.io/beads/ as the upstream home. Older steveyegge/beads links may redirect.
• Default to plain bd ... commands in 1.0+. Do not assume pre-1.0 patterns like bd --no-daemon ..., bd sync, or SQLite-backed storage still apply.
• Use --json for machine parsing where supported, and prefer non-interactive flags over editor-based flows.
• Use bd ready, bd show <id>, bd update <id> --claim, bd dep add, and bd close <id> --reason "..." for the core issue lifecycle.
• Use bd onboard for the minimal snippet that belongs in AGENTS.md or equivalent agent instructions files.
• Use bd prime as the single source of truth for session workflow and context recovery after compaction or agent restart.
• Prefer bd bootstrap for fresh clones or broken local setup, and reserve bd init --force or bd doctor --fix for explicit repair work after reviewing the impact.
• Treat bd setup ... --print output as an install template, not the authoritative command reference.