doey

Plan-first task creation — research, breakdown, risk analysis, then create tasks from approved plan. Usage: /doey-planned-task <goal>

Request a respawn of the current worker pane. Use when you need to "respawn me", "restart myself", "fresh context", or "reset my pane". Writes an atomic respawn request that kills the current Claude process and relaunches with a fresh context.

Discover scaffolding patterns in this project — structural fingerprints, accretion files, and refactoring patterns from git history. Returns candidates worth turning into templates.

Spawn the doey-scaffy MCP server so workers can drive the Scaffy template engine via tool calls. Loaded automatically when a task needs scaffy_run / scaffy_validate / scaffy_list / scaffy_audit / scaffy_discover / scaffy_new / scaffy_fmt or any scaffy:// resource.

Create a new scaffy template from existing files (--from-files) or interactively. Outputs to .doey/scaffy/templates/<name>.scaffy.

Manage persistent project tasks — list, add, transition, show, subtasks, decisions, notes. Tasks stored in .doey/tasks/ (survives reboot). Use when you need to "show tasks", "add a task", "mark a task done", "what are we working on", or "create a task".

Quick inline clarification loop — a few targeted AskUserQuestion rounds to sharpen a vague goal before task creation. Use when you need to "clarify", "interview the user", "sharpen a vague goal", or resolve ambiguity without spawning a full /deep-interview window.

Spawn a team from a .team.md definition file. Usage: /doey-add-team <name>

Add a new team window (Subtaskmaster + Workers), optionally as a reserved freelancer pool.

Quick task creation without planning — create and dispatch immediately. Usage: /doey-instant-task <goal>

Kill a team window — stop processes, remove tmux window, clean runtime files.

List all team windows with their status. Use when you need to "show team windows", "list teams", or "what teams are running".

Strategic planning with a mandatory interview phase followed by ultrathink research and vertical phase design. Usage: /doey-masterplan [--quick] <goal>

Spawn a Doey R&D product team — Brain, Platform Expert, Claude Expert, Critic. Works on the live codebase so it sees all changes in real time.

Kill and relaunch Claude instances. Use when you need to "restart workers", "reset the team", "clear and relaunch", or "fresh start". Resets process, context, name, agent, status. Skips reserved workers unless --force.

Send tasks to idle worker panes. Use when you need to "send a task", "assign work", "delegate to workers", "dispatch to idle panes", or "give a worker something to do".

Open an interactive Settings window with a config editor and live settings panel. Use when you need to "open settings", "configure doey", or "change doey settings".

Isolate a team window in a git worktree, or return it.

Full codebase simplification across all teams. Taskmaster inventories capacity, assigns domains, dispatches to Subtaskmasters. Use when you need to "simplify the whole codebase", "reduce complexity everywhere", or "run a codebase-wide cleanup".

Compile a natural-language goal into a structured task package (.task + .json). Usage: /doey-create-task <goal>

Dispatch a research task to a worker. Stop hook blocks until report is written. Use when you need to "research a topic before implementing", "investigate an approach", or "spawn a research agent".

Spawn a Deep Interview window — structured requirements extraction before complex tasks

Diagnose and repair Doey Dashboard (window 0). Use when you need to "fix the dashboard", "repair window 0", or "dashboard panes are broken".

Toggle debug flight-recorder mode for Doey hooks. Use `/doey-debug on` to start recording, `/doey-debug off` to stop, `/doey-debug status` to view state. Captures hook timing, state transitions, lifecycle events, and IPC messages.

Kill ALL Doey tmux sessions, processes, and runtime files. Use when you need to "kill all doey sessions", "shut down everything", or "nuke all doey instances".

Audit Doey's multi-pane context for token waste and bloat. Use when user says: audit context, audit setup, why is doey slow, token waste, context bloat, audit the team, or invokes /doey-context-audit. Scans all active panes, agent definitions, CLAUDE.md, skills, hooks, msg queue pollution, task file bloat, duplicate rule text. Returns team-level health score with auto-fix and diff-confirm flow.

On-demand snapshot of worker pane state — status files (FINISHED, BUSY, ERROR, READY, RESERVED), reservations, results, crashes. Use ONLY when you explicitly need a one-shot picture of the team. Never use as a background watch loop.

Audit scaffy templates for staleness with 6 checks — anchor validity, guard freshness, path existence, variable alignment, pattern activity, structural consistency. Returns healthy / needs_update / stale.

Apply a scaffy template to the current project. Supports --var KEY=VALUE, --vars-file, --dry-run preview, --diff unified diff output, --json structured report. Use /doey-scaffy-run <template> [vars].

Validate a scaffy template syntax and structure. Use --strict to also flag missing transforms, anchors, IDs.

Example skill demonstrating MCP server integration. Not for actual use.

Reserve/unreserve the current pane to prevent Subtaskmaster dispatch. Use when you need to "reserve this pane", "prevent dispatch to my pane", or "unreserve a worker".

View or set pane status for Doey workers. Team-wide view with `/doey-status team`. Use when you need to "check worker status", "see which panes are busy", or "set a pane to READY".

Hot-reload Subtaskmaster without stopping workers. Use when you need to "reload doey config", "restart subtaskmaster", or "apply agent/hook changes".

Cascading nudge — wake stalled Claude instances across all teams. Use when workers seem stuck, idle, or unresponsive.

Stop a worker by pane number. Use when you need to "stop a worker", "kill a pane", or "halt a task". Kills Claude process, updates status.

Factory reset Doey — restore default settings, reset config and agents. Use when you need to "factory reset doey", "restore defaults", "reset config and agents", "reset doey to defaults", or "reset doey settings".

Kill the entire Doey session — all windows, processes, and runtime files. Use when you need to "kill this session", "shut down doey", or "stop everything and clean up".

Two-wave purge — audit every project file for context rot and code issues, then fix. Use when you need to "audit the codebase", "clean up context rot", or "purge stale content".

Reinstall Doey from source repo. Use when you need to "reinstall doey", "update doey from source", or "refresh doey installation".