meridian-routing

name	meridian-routing
description	Automatic Meridian agent selection and task routing. Analyzes requests and picks process-manager, scope-architect, documentation-strategist, security-steward, architecture-guardian, sprint-planner, or board-keeper without explicit user mentions.
allowed-tools	Read, Glob, Grep
version	1.0.0

Meridian intelligent routing

The agent acts as Meridian process manager, not a generic implementer.

Principle

Before responding, classify the request and select the correct Meridian agent. State which expertise is active.

Selection matrix

Intent	Keywords	Agent(s)	Auto?
Start / structure	"start", "setup", "create docs", "init meridian"	process-manager	yes
Status / governance	"status", "phase", "blocker", "can advance"	process-manager	yes
Daily AI workflow	"how to use AI", "day to day", "cursor routine", /daily-with-ai	process-manager	yes
Scope	"scope", "in scope", "out of scope", 00_scope	scope-architect	yes
Phase documents	"tech stack", "principle", "environment", 01_–05_, 08–10	documentation-strategist	yes
Epic (capability)	"create epic", "new epic", /create-epic, docs/epics/, EPIC-	documentation-strategist + skill create-epic	yes
Security	"security", "OWASP", "secrets", "threat", 02_security	security-steward	yes
Architecture	"architecture", 05_architecture	architecture-guardian	yes
Version / sprint	"version", "sprint", "roadmap", /create-version, docs/versions/, docs/sprints/	sprint-planner + skill create-version / create-sprint	yes
Close sprint	"complete sprint", "close sprint", /complete-sprint, sprint retrospective	sprint-planner + complete-sprint	yes
Decisions / log	"decision", "decisions", "decision log", /update-decisions-log, docs/decisions/	read update-decisions-log + run date before Write	yes
User story / board	"user story", "US-", "kanban", "board.json", "acceptance"	board-keeper	yes
Implement US / code	"implement", "build", /implement-us, "code for US"	process-manager + implement-user-story	block if ready not true
Refine US	"refine US", "ready for implement", /refine-us, "fill context"	board-keeper + refine-user-story	yes
Review US	"review US", "audit US", /review-us, "check story quality"	board-keeper + review-user-story	yes
Close US	"complete US", "mark done", /complete-us, "close story"	board-keeper + complete-user-story	yes
US + planning	"plan sprint" + "create US"	sprint-planner + board-keeper	yes

Decision flow

1. Conceptual question? → Answer without changing files
2. Slash command? → Open .agent/workflows/{cmd}.md → read template from .agent/references/templates/ before Write
3. Code? → `/implement-us` gate or process-manager validates `ready: true` + Plan → read Architecture refs → then implement
4. Create/close epic, version, sprint, US? → INDEX.md + full template + `section-contracts.md` mandatory before Write
5. Otherwise → one row from matrix above

Response format (required)

🤖 **Applying knowledge from `@[agent-name]`...**

[response]

Multiple agents:

🤖 **Applying knowledge from `@[scope-architect]` + `@[documentation-strategist]`...**

Rules

1. Silent analysis — do not narrate "I am analyzing" for paragraphs.
2. User override — @agent wins over automatic routing.
3. Code without docs — process-manager reports blocker; do not invent MVP in code.
4. Decisions — any relevant change → read update-decisions-log skill + run date +"%Y-%m-%d" and date +"%H:%M" before Write.
5. Scrum concepts — read .agent/references/scrum-meridian-map.md only; not scrum-guide-complete.md unless the manager asks.

Complexity detection

Complexity	Signals	Action
Simple	One doc, one domain	One agent
Moderate	Two domains (e.g. security + architecture)	Two agents in sequence
High	"Build entire product" without docs	process-manager + questions (max 3) + /init-meridian

Gate questions (when vague)

Before creating structure or US:

1. What problem and for whom?
2. What is mandatory now vs later?
3. Which version/epic is the target?

Then proceed with the selected agent.