router

name	router
description	Use this skill when you are unsure which Ravie skill applies, need to sequence multiple workflows, or need a human-readable routing reference for a mixed request. Do not use when the request clearly matches one specific skill; use that skill directly.

router

Routing reference for when no single skill clearly matches. This is a human-readable map, not automatic routing logic.

Note: This skill contains an abbreviated quick-reference routing map. For the full route list across all 33 skills, see ROUTER.md at the repo root.

Purpose

When you are unsure which skill applies, determine: what type of work is this, which skill should be read next, what context must be gathered first, and what permission tier applies.

When to use this

Use this skill when the request matches the YAML description and you need to choose among Ravie workflows.

When NOT to use this

• Do not use when the request clearly matches one specific skill; use that skill directly.

Routing table

Request pattern	Route to	Gather first
"Build feature X" / "Implement Y"	requirements-griller → issue-to-pr	Linear issue, acceptance criteria
"I have an idea for..." / "What if we..."	requirements-griller deep interrogation mode	Nothing — start deep interrogation mode
"Fix bug" / "This is broken" / "Error in..."	debug-root-cause	Error message, reproduction steps
"Create issues from PRD" / "Plan this feature"	idea-to-prd-tracer	Notion PRD or design file
"Implement this design" / "Build from Figma"	figma-lovable-handoff → issue-to-pr	Design file/URL
"What should I work on?" / "Morning brief"	daily-brief	Nothing — skill pulls from systems
"Review this code" / "Check before merging"	code-review	Diff or branch name
"Deploy" / "Ship it" / "Is this ready?"	deploy-ready	Branch, preview URL
"Log this decision"	decision-log-adr	The decision and context
"Create Linear issues" / "Update status"	linear-operator	Issue details
"Update Notion" / "Save this to Notion"	notion-brain	Content to save
"Check the VPS" / "Job failed"	automation-sre	Job name, error
"Launch plan" / "Positioning" / "Competitors"	growth-launch-pack	Product context
"Something is down" / "Production error"	observability-incident-loop	Error evidence
"Set up Claude Code for this project"	project-control-plane	Repo path

System of record routing

Data type	System	Skill
Durable knowledge (PRDs, decisions, runbooks)	Notion	notion-brain
Execution (tasks, status, priorities)	Linear	linear-operator
Code (source, PRs, CI, releases)	GitHub	github-operator
Database/auth/storage	Supabase	supabase-guardian
Preview/deploy	Vercel	vercel-preview-qa
Automation/scheduled jobs	VPS	automation-sre

Permission classification

Before executing any routed action, classify the permission tier:

• Reading from any system → Tier 1 (auto)
• Drafting content locally → Tier 2 (auto, announce)
• Writing to external systems → Tier 3 (ask first)
• Destructive or production actions → Tier 4 (name the target)

Routing rules

1. If a skill exists for the task, use it. Don't improvise when there's a defined workflow.
2. If the request is ambiguous, ask one clarifying question — don't guess.
3. If multiple skills apply, route to the first one in the chain (e.g., requirements-griller deep interrogation mode before scope finalization, then issue-to-pr).
4. If no skill matches, handle directly but flag that a skill might be worth creating.

Hard rules

• Never skip a skill that matches the task
• Never guess which system to write to — check the routing table
• When uncertain about routing, ask the user rather than picking the wrong skill
• Every external action goes through permission classification first

Output format

# Routing recommendation

## Request
[The user's request, summarized]

## Recommended skill
`[skill-name]`

## Reason
[Why this skill matches the request]

## Required context
[What context to gather before invoking the skill]

## Permission tier
Tier [1-5] — [classification reason]

## Next action
[Specific first step the user or Claude should take]

## Alternative skills considered
- `[skill]` — rejected because [reason]

Connects to

• All 33 skills — this is the routing index
• permission-guardian — for permission tier classification
• system-of-record-governance — for system-of-record routing decisions

Common failure modes

Over-routing — Routing to a skill when a 2-line direct answer would work. Skills are for repeated, structured workflows, not every request. If the answer is "run X", just say so.

Skipping permission classification — Routing to a skill that does external writes without flagging the permission tier. Every routing decision must include the permission classification.

Choosing implementation before requirements — Routing to issue-to-pr when scope isn't clear. Always check if the request has acceptance criteria first. If not, route to requirements-griller regardless of how implementation-shaped the request looks.

Picking the wrong skill from the chain — When multiple skills apply, route to the earliest in the chain. Idea → grill (deep mode of requirements-griller) → requirements-griller → idea-to-prd-tracer → issue-to-pr. Don't jump to issue-to-pr if requirements aren't done.