next-move

license	Apache-2.0
name	next-move
description	Predicts the highest-impact next action for your project by running a 5-agent meta-DAG pipeline. Gathers project signals automatically (git, recent files, port-daddy, CLAUDE.md), then runs sensemaker → decomposer → skill-selector + premortem → synthesizer. When execution is approved, convert each predicted node into a skillful node prompt using skillful-node-prompt + skillful-subagent-creator, prefer live WinDAGs visualization backed by POST /api/execute and /ws/execution/:id, and fall back to ASCII only when live visualization is unavailable. Activate on: "what should I do", "what's next", "next move", "/next-move", "where should I focus", "what's the highest impact thing right now". NOT for: creating skills, debugging one specific bug, or promising topology-specific runtime behavior the current server cannot execute.
allowed-tools	["Read","Grep","Glob","Agent","Bash(git:*)","Bash(pd:*)","Bash(find:*)","Bash(ls:*)","Bash(head:*)","Bash(cat:*)","Bash(wg:*)","Bash(node:*)","Bash(pnpm:*)","Bash(open:*)","Bash(curl:*)","mcp__windags__windags_skill_search","mcp__windags__windags_skill_graft","mcp__windags__windags_skill_reference","mcp__windags__windags_history"]
metadata	{"category":"Agent & Orchestration","tags":["planning","decision-making","meta-dag","skill-selection","context-analysis","sub-agents","live-execution"],"pairs-with":[{"skill":"windags-sensemaker","reason":"Problem classification and halt gate"},{"skill":"windags-decomposer","reason":"Three-pass decomposition protocol"},{"skill":"windags-premortem","reason":"Failure pattern scanning"},{"skill":"task-decomposer","reason":"Breaks recommended moves into executable subtasks"},{"skill":"skillful-node-prompt","reason":"Defines the 4-branch execution prompt structure for every node"},{"skill":"skillful-subagent-creator","reason":"Defines role boundaries, equipped skills, and node input/output contracts"},{"skill":"output-contract-enforcer","reason":"Downstream nodes need explicit structured outputs, not vibes"},{"skill":"human-gate-designer","reason":"Approval nodes should become actual gates, not free-text TODOs"},{"skill":"reactflow-expert","reason":"Live DAG visualization should match execution state, not a separate mock"}],"argument-hint":"[--fresh] [focus hint]","user-invocable":true}

/next-move

Predict the best next move, present it clearly, and — on approval — hand it to the real WinDAGs execution path.

Two operating rules are mandatory:

1. Separate planning topology from runtime topology. Server execution is real for dag and workflow. team-loop, swarm, blackboard, team-builder, and recurring fall back to generic DAG execution. See references/runtime-honesty.md.

2. Prefer live WinDAGs visualization over ASCII when execution is going to happen. ASCII is fallback only. See references/live-execution-visualization.md.

Arguments: $ARGUMENTS

If the user passed --fresh, ignore conversation history and predict from project signals only.

Skill Layout

This skill is structured for on-demand loading. Don't bulk-load — pull only what the current stage needs.

Need	Load
Decision tree, file map, top-level guidance	INDEX.md
Per-agent prompts (Sensemaker, Decomposer, etc.)	prompts/<agent>.md
Operational rules and judgment-call guidance	references/<topic>.md (see references/INDEX.md)
Worked end-to-end walkthroughs	examples/<NN>-<scenario>.md (only when shape recognition is needed)
Output schemas for validation	schemas/<output>.schema.json
Empty starting structures	templates/<output>.template.<ext>
Replay / validate / inspect-attribution utilities	scripts/<utility>.{sh,ts}
Direct invocation of a single meta-DAG stage	agents/<stage>.md (subagent for Task tool)

Project Signals

These are preprocessed at skill-load time via !... shell calls. Treat them as ground truth.