Skill for using the AgentBoard AI project management app and its cloud-hosted MCP server. Use when the user wants to manage phase-based projects (phases, milestones, documents, tasks), work with workspace boards (apps, boards, cards, artifacts) for ad-hoc orchestration, authenticate the AgentBoard MCP, interact with any AgentBoard MCP tool, or needs guidance on either workflow. Triggers on "manage project", "use agentboard", "agentboard workflow", "create a project in agentboard", "set up agentboard", "authenticate agentboard", "create a workspace card", "create a board", "orchestrate cards", "submit artifact", "workspace board", or "fetch a card".
Convert an approved AgentBoard spec into an architecture document and workspace cards — runs the level-aware pipeline (research → classification audit → L1/L2/L3 compose → design review), surfaces the bundle, audit, level, and design-review findings as transparency, then commits the architecture document and creates one card per Card Slice. Use ONLY when the user explicitly wants to run the architecture phase on an existing approved spec — e.g. "run architecture on docs/specs/X", "turn the spec into cards". Do NOT trigger on general architecture/design discussion, or when there is no approved spec yet (use foundation first).
Read-only progress snapshot of an AgentBoard workspace board — card counts per column, percent finished, blocking settings, cards needing attention, and recent activity, without resuming work. Use whenever the user asks how a board or orchestration is going, where a board stands, how many cards are left or finished, or wants a board snapshot — e.g. "board status", "how's the cleanup board doing", "how many cards left on the board", "show me where the board is at". Mutates nothing.
Run an interactive spec-building session for AgentBoard — produce a rigorous, architecturally-silent spec via the spec-writing skill, commit it, and hand off to architecture. Use whenever the user wants to start a new body of work that needs a spec before architecture and cards — e.g. "foundation", "write a spec for this feature", "let's spec this out", "start the foundation phase". Produces docs/specs/<file>.md; it does NOT make architecture decisions or create cards (that is the architecture skill).
Onboard a fresh agent to AgentBoard and start a project: authenticate, health-check, list or create a project, and claim the first task. Use whenever the user wants to start, set up, kick off, or begin a new AgentBoard project, or hand off to a brand-new session — e.g. "kickoff", "start a new agentboard project", "set this project up in agentboard", "onboard a new agent". To resume existing work instead, use pickup.
Run the AgentBoard workspace pipeline across a board's cards — planning → review → implementation → audit waves with parallel subagents per card, server-driven verdict routing, and checkpoint pauses. Use ONLY when the user explicitly wants to run or continue the pipeline on a board that already has cards — e.g. "/orchestrate", "run the pipeline", "orchestrate the board". Do NOT trigger on general discussion of orchestration, or when no cards exist yet (create them via architecture or sweep first).
Resume existing AgentBoard work in a continuation session: authenticate, find the active project, resume any in-progress task or claim the next one, and read recent activity for context. Use whenever the user wants to pick up, resume, continue, or get back into an existing AgentBoard project — e.g. "pickup", "resume agentboard", "continue where we left off", "what was I working on". To start a brand-new project instead, use kickoff.
Read-only status of an AgentBoard phase-based project — current phase (N/13), task counts by status, blockers, and the next action, without starting work. Use whenever the user asks where a project stands, what phase it's on, what's blocked, or what's next on an AgentBoard project — e.g. "project status", "where are we", "what phase are we in", "anything blocked". Mutates nothing. For a workspace-board snapshot, use board-status instead.