The aiguide MCP server ships an aiguide_install_skill tool that gives you the SKILL.md contents and the install command in one call:

Install the morning-brief skill from aiguide

Claude calls aiguide_install_skill { slug: "morning-brief" }. The response includes:

• The full SKILL.md body (description, trigger words, what tools to call, what to output)
• A copy-paste install command that drops it into ~/.claude/skills/morning-brief/SKILL.md
• A short summary of what the skill does

Run the copy-paste command in your shell. Restart Claude Code. The skill is registered.

If you do not have aiguide connected, the file lives in this guide's source — copy it from there. Or write your own (Step 6).

A morning-brief skill is mostly orchestration of tools you already have. The default version calls these in parallel:

• nex_sprint — current sprint, blocked items, in-progress, recently shipped
• nex_proactive — stale learnings, open decisions, knowledge gaps, pattern clusters
• (optional) mcp_fleet_health.check_all_servers — if you run a fleet of services
• (optional) crm_follow_up list — if you have a CRM with follow-ups
• (optional) n8n_list_executions failed — if you have n8n workflows

The skill bundles them into one parallel call so the brief lands in 2-3 seconds, not 10. Cross-domain check: code + system + business + backlog all on one screen.

If you are solo and not running a server fleet, drop the fleet-health step. If you do not have a CRM, drop the follow-up step. The skill is just a Markdown file at ~/.claude/skills/morning-brief/SKILL.md — open it, delete the lines for tools you do not have.

Default trigger phrases (any of these work):

• "morning brief"
• "what do I need to know today"
• "status check"
• "/morning-brief" (if you also have a slash command)

Keep the trigger list short and natural. The model picks up the skill when any of those phrases appear.

Run aiguide_validate_step. The validator checks ~/.claude/skills/ exists and contains at least one entry. After you install morning-brief that is automatic.

Tomorrow at 09:00:

Morning brief

Output (compact, one screen):

Morning brief — Friday 25.04
─ Sprint: 2 in-progress (deploy aiguide v0.3, fix subagent flake), 1 blocked (waiting customer feedback)
─ Open decisions: RLS migration (parked since Mon), pricing-tier B/C consolidation
─ Stale learnings: 4 (3 about Postgres pool sizing — synthesize candidate)
─ Health: memory ok, 0 contradictions, 1 dedupe pending
─ Followups today: ping customer-A on contract, review @max PR-pitch draft
─ n8n: 1 failed run last 24h (welcome-mail to test-foo@... — already triaged)

That is the entire day's framing in 6 lines. If something on it is urgent, you do that first. If nothing is urgent, you go to sprint top item with confidence that nothing important is bleeding behind your back.

The default skill assumes the StudioMeyer-style stack. If your stack is different, copy the SKILL.md and edit. The pattern:

---
description: Morning brief — what do I need to know today. Aggregates sprint, health, follow-ups, failures across my stack. Cross-domain decision frame for the day.
---

When the user says "morning brief", "what do I need to know today", or "/morning-brief":

1. Call these in parallel (one tool-use block, multiple tools):
   - <list your tools here>

2. Aggregate into 6-10 lines max. One section per domain.

3. Lead with anything that is urgent. End with what is on sprint top.

4. No emojis. No headers. Just dense factual lines.

The 6-10 line constraint is the most important rule. A morning brief that is two screens is a daily report, not a brief, and you stop reading it.

What this is NOT

Not a daily standup template. Not a calendar replacement. Not a project dashboard. Morning brief is the answer to "what should I worry about right now?" — not "what is the full state of my world?". Keep it short, factual, and runnable in 30 seconds. The full state lives in your project tools, not in the brief.