task-management

Plugin task-management from marketplace bytedesk in ByteDeskAI/bytedesk-marketplace. Commands read from the repo's manifest — 2 days ago.

Skills (27)

• adr

  Write an architecture decision record into .bytedesk/task-management/adrs/, or finish one of the stubs the AskUserQuestion hook drafted automatically. Use when a real decision gets made — a tradeoff, a rejected alternative, a constraint the team must not relitigate — or when the user says "record th

• agent

  Dispatched-worker registry — list with derived liveness, heartbeat a name, reap the dead (parks their claimed tasks). Use when the user says "who is running", "reap dead workers", "agent heartbeat", "/agent", or a dispatched claim looks abandoned.

• board

  Show the task board for this repo — epics, in-progress, blocked, next unblocked, and stale work — from .bytedesk/task-management/. Use when the user asks "what's on the board", "where are we", "what's next", "/board", or when resuming work and you need the real state rather than session memory.

• caps

  Probe what this host can dispatch work to — topology, orchestration, tmux, manual, harness CLIs, sandbox binaries — via tm caps / GET /api/caps. Use before tm dispatch or tm pool, when a backend is "unavailable", or when the user says "what can this machine run", "host capabilities", "/caps".

• collect

  Pull a dispatched worker's result into the store — never closes the task; parks and releases on failure. Use when a worker has exited, after tm dispatch / tm pool, when the board still shows in_progress for a dead run, or the user says "collect TM-N", "what happened to that worker", "/collect".

• dispatch

  Hand one ready-for-agent task to a worker end to end — claim, start, worktree, handoff, spawn (topology → tmux → orchestration → manual). Use when the user says "dispatch this", "run this on an agent", "spawn a worker for TM-N", "/dispatch", or a labelled card should execute without this session doi

• enhance-capture

  Snapshot what the product actually does today into .bytedesk/task-management/product-state.md, so capability proposals are grounded in the repo rather than in session memory. Use as step 1 of /enhance, or when the user asks to refresh product state or capability context.

• enhance-mine

  Mine what already happened — Claude transcripts, the board, pool.log and test logs — for recurring errors, workarounds and user complaints, then file bugs as tasks and enhancements as CAP proposals, deduped against the board. Use when the user runs /enhance-mine, says "mine the transcripts", "what k

• enhance-propose

  Turn research into ranked capability cards in the task store (.bytedesk/task-management/bin/tm cap new), each sized by impact/effort/confidence with checkable acceptance criteria. Use during /enhance or when the user asks for enhancement ideas, a product backlog, or "what should we build next".

• enhance-research

  Research for the capability loop — internal code seams plus external analogues and competitors — written to a dated pack under .bytedesk/task-management/research/. Use for the /enhance research step or a standalone "research similar products" / "what do comparable tools do" request.

• enhance-track

  Move capabilities through their lifecycle with evidence — accept a proposal into a task, mark one shipped, or drop it. Use post-merge, when the user says "we shipped CAP-X", when auditing which proposals actually landed, or as the closing step of /enhance.

• enhance

  Product capability loop — capture current product state, research external analogues and internal seams, propose ranked capability cards into the task store, and track them to shipped. Use when the user runs /enhance, says "enhance the product", "what should we build next", "capability backlog", or

• epic

  Open or switch the active epic for this repo. Every task must belong to an epic — the PreToolUse gate denies TaskCreate until one is active, so this is the front door for any new body of work. Use when the user says "start an epic", "/epic", "new initiative", "switch epics", or when a TaskCreate was

• events

  Raw task-store event stream — tm events [--json] [--follow] [--since], webhooks, dashboard SSE. Use when the user says "tail the board events", "webhook the store", "follow JSONL", "/events", or a machine (not a person) needs what tm log renders for humans.

• groom

  Backlog grooming pass over .bytedesk/task-management/ — close zombies, merge duplicates, unstick blocked tasks whose blockers are done, fill missing acceptance criteria, and re-rank what's next. Use when the board has drifted, before planning a new epic, or when the user says "clean up the backlog",

• handoff

  Produce a self-contained brief for one task — context, acceptance criteria, blockers, evidence, commits — so a subagent, a worktree session, or tomorrow's session can pick it up cold. Use when delegating work, spawning a parallel agent, ending a session mid-task, or when the user says "hand this off

• implement

  Build one implementation ticket — claim, verify, attach evidence, then submit governed work for review or close ungoverned work. Use when a ready-for-agent task is startable, or the user says /implement.

• interview

  Relentless HITL interview that records decisions on a task (and ADRs when they are hard to reverse). Resolves decision:interview tickets. Use for /interview, grilling a plan, sharpening a destination, or single-session planning that is not big enough for /map.

• map

  Chart a foggy, multi-session effort as a decision map on the task store — an epic labelled decision:map whose children are interview/research/prototype/unblock tickets — then work the frontier one ticket at a time until the way is clear. Use when the destination can be named but the route cannot, or

• override

  Bypass the next task-management gate once (no active epic, WIP limit, unmet acceptance criteria, or a blocking Stop), with the reason recorded in the audit log. Use when a gate is wrong for the situation and the user wants through anyway — "override", "skip the gate", "/override", "just let me creat

• pool

  The pickup loop that dispatches ready-for-agent tasks, on by default, as one detached pool per repository — tm pool once|ensure|start|stop|status|resume|wait, the tm-pool monitor, dispatch.enabled / poolWip / pollSeconds / maxFailures / idleExitMinutes. Use when the user says "start the worker pool"

• prototype

  Throwaway prototype that answers one design question (look or behave). HITL — a human picks the variant. Resolves decision:prototype tickets. Use for /prototype, "how should this look", "does this state model feel right".

• research

  AFK research against primary sources; write a cited pack under .bytedesk/task-management/research/ and attach it as evidence. Resolves decision:research tickets. Use when a fact outside the working directory is blocking a decision, or the user says /research.

• route

  Pick which task-management flow fits — board, enhance, interview, map, spec, tickets, implement, or the agent-first loop (caps → dispatch → pool → collect → events) — from live store state. Use when the user is unsure how to start.

• spec

  Collapse a decision map or the current thread into a buildable spec in .bytedesk/task-management/plans/ and set epic.plan. Use after a map is clear, or when deciding is already done. Do not interview; synthesise.

• standup

  Summarize what changed in the task store since a point in time — tasks opened, started, closed, blocked, evidence and commits attached — read from the append-only event log. Use for "what did we do yesterday", "standup", "status update", "/standup", or when writing a progress report for someone who

• tickets

  Break a spec or conversation into tracer-bullet implementation tasks with acceptance criteria and blockedBy edges. Use after /spec, or when a plan is ready to slice. Never put decision:* labels on these cards.

MCP servers (1)

• task-management

  runs: ${CLAUDE_PLUGIN_ROOT}/bin/tm-mcp

Hooks (15)

• SessionStart

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh session-start

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh pre-task-create

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh pre-bash

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh post-task

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh post-plan

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh post-bash

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh post-edit

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh post-decision

• UserPromptSubmit

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh user-prompt

• PreCompact

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh pre-compact

• SubagentStart

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh subagent-start

• SubagentStop

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh subagent-stop

• Notification

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh notification

• SessionEnd

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh session-end

• Stop

  ${CLAUDE_PLUGIN_ROOT}/hooks/tm-hook.sh stop

Similar plugins