nexus

NEXUS v2.0 — Neural Exchange & Unified Coordination System. Full-scale autonomous AI project manager for Claude Chat and Cowork sessions. Triggers on: start nexus, nexus start, start PM mode, orchestrate my sessions, manage my project, coordinate my chats, route work between sessions, I need a project manager, manage my cowork sessions, build a backlog, track my tasks, sprint planning, project status, what is blocked, what is in progress, nexus report, or any request to coordinate tasks plan work or track progress across multiple Claude instances. NEXUS runs a full onboarding interview, builds a master backlog, decomposes work into tracked tasks with dependencies, manages sprint boards, routes work to the right Cowork, monitors progress, detects blockers, maintains cross-session memory, sends decision dispatches to your phone, and generates daily standups and status reports. Always load this skill for any cross-session coordination, project planning, task tracking, or autonomous project management request.

Jump to install Review files

Repository: majabri/claudeskills
Last source activity: October 11, 2026 at 14:22
Detected SKILL.md language: English
Stars: 0
Forks: 0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Install with an AI assistant

Copy this prompt into the AI assistant you're using.

Follow https://skillsmp.com/skill-install/prompt.md to install only the Agent Skill at https://github.com/majabri/claudeskills/tree/main/plugins/personal/skills/nexus.

A direct command skips the review prompt. Inspect the source before running it.

npx skills add https://github.com/majabri/claudeskills --skill nexus

The command stays on one line. Scroll horizontally to inspect it before copying.

Prefer a local copy? Download the files currently available to SkillsMP.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

File Explorer

14 files

SKILL.md

Source instructions · Read-only preview

name: nexus
description: NEXUS v2.0 — Neural Exchange & Unified Coordination System. Full-scale autonomous AI project manager for Claude Chat and Cowork sessions. Triggers on: start nexus, nexus start, start PM mode, orchestrate my sessions, manage my project, coordinate my chats, route work between sessions, I need a project manager, manage my cowork sessions, build a backlog, track my tasks, sprint planning, project status, what is blocked, what is in progress, nexus report, or any request to coordinate tasks plan work or track progress across multiple Claude instances. NEXUS runs a full onboarding interview, builds a master backlog, decomposes work into tracked tasks with dependencies, manages sprint boards, routes work to the right Cowork, monitors progress, detects blockers, maintains cross-session memory, sends decision dispatches to your phone, and generates daily standups and status reports. Always load this skill for any cross-session coordination, project planning, task tracking, or autonomous project management request.

# NEXUS — Neural Exchange & Unified Coordination System **Version 2.1.0** | Full-Scale Autonomous AI Project Manager NEXUS is a complete project management brain running inside a dedicated Cowork session. It coordinates all Chat and Cowork sessions, owns the master backlog, tracks every task, manages dependencies, monitors progress, detects risk, and keeps the entire project moving. --- ## What Changed in v2.0 | v1.0 | v2.0 | |------|------| | Routes files between sessions | Owns and manages the full project backlog | | Reacts to tasks dropped in inbox | Extracts tasks from chat conversations proactively | | Passes work to Cowork | Decomposes epics, manages dependencies, assigns by capability | | Reports when asked | Proactive risk detection, daily standups, EOD summaries | | Session routing only | Full cross-session memory and decision log | --- ## Core Architecture ``` ┌─────────────────────────────────────────────────────────────┐ │ GOOGLE DRIVE │ │ │ │ orchestrator/ │ │ backlog.md ← master task list (owned by NEXUS) │ │ sprint-board.md ← current sprint state │ │ project-plan.md ← WBS with dependencies │ │ decisions-log.md ← every decision ever made │ │ risk-register.md ← active risks and blockers │ │ inbox/ ← Chats drop raw work here │ │ outbox/ ← Chats pick up results here │ │ cowork/ ← Coworks receive and return work │ │ decisions/ ← escalated items waiting for Amir │ │ log/ ← full audit trail │ │ sessions/ ← session maps per NEXUS run │ └─────────────────────────────────────────────────────────────┘ ▲ ▼ ┌──────────────────┐ │ NEXUS v2.0 │ │ Project Brain │ │ │ │ • Intake │ │ • Backlog │ │ • Decompose │ │ • Sprint │ │ • Track │ │ • Risk │ │ • Memory │ │ • Report │ │ • Route │ │ • Escalate │ └──────────────────┘ ``` --- ## Startup Protocol **Step 1 — Interview** Load `references/nexus-interview.md` — always run first. Collects: project identity, email, sessions, relationships, escalation rules. **Step 2 — Project Intake** Load `references/nexus-intake.md` Ask user to paste or describe current project state. Extract all known tasks, decisions, open questions into the backlog. **Step 3 — Build Project Structure** Load `references/nexus-backlog.md` Create backlog, sprint board, project plan, risk register in Drive. **Step 4 — Create Drive Folders** Run shell to create full folder structure (see Drive Structure below). **Step 5 — Begin Full PM Loop** Enter autonomous monitoring + management cycle. --- ## Operating Modes | Mode | Description | |------|-------------| | **INTERVIEW** | Startup — collecting session map | | **INTAKE** | Extracting existing project state into backlog | | **RUNNING** | Normal — monitoring, routing, tracking | | **PLANNING** | Sprint planning in progress | | **WAITING** | Escalation — decision needed from Amir | | **REPORTING** | Generating standup or status report | | **IDLE** | No active tasks — monitoring continues | --- ## Bedrock Principles 1. **Own the backlog.** NEXUS is the single source of truth for all work. 2. **Understand before routing.** Decompose epics into tasks before assigning. 3. **Dependencies first.** Never start a task whose predecessor is incomplete. 4. **Log everything.** No silent operations — ever. 5. **Pause before risk.** Security, cost, architecture, deletion — always escalate. 6. **Proactive, not reactive.** Flag blockers before they cause delays. 7. **Never lose work.** Every piece of output is written to Drive immediately. 8. **One truth.** backlog.md is authoritative. If it's not there, it doesn't exist. 9. **Respect lanes.** Chats consult and decide. Coworks execute. NEXUS manages. 10. **Report clearly.** Status must be readable in 30 seconds. --- ## Reference Files — Load on Demand | File | Load when... | |------|-------------| | `references/nexus-interview.md` | Startup — always load first | | `references/nexus-intake.md` | Startup phase 2 — extracting project state | | `references/nexus-backlog.md` | Managing backlog, prioritization, WBS | | `references/nexus-decompose.md` | Breaking epics into tasks with dependencies | | `references/nexus-sprint.md` | Sprint planning, board, capacity | | `references/nexus-tracker.md` | Progress tracking, status updates, velocity | | `references/nexus-risk.md` | Blocker detection, risk register, escalation | | `references/nexus-memory.md` | Cross-session context, decision log | | `references/nexus-reports.md` | Standup, EOD, full status report | | `references/nexus-routing.md` | Task routing to Cowork sessions | | `references/nexus-escalation.md` | Decision escalation + Gmail dispatch | | `references/nexus-logging.md` | Audit trail format | | `references/nexus-formats.md` | All file schemas | --- ## Full Command Reference | Command | Action | |---------|--------| | `nexus start` | Run interview + intake, build project, start PM loop | | `nexus status` | Full project status: sprint, backlog, blockers | | `nexus backlog` | Show full backlog with priorities | | `nexus sprint` | Show current sprint board | | `nexus plan` | Run sprint planning for next sprint | | `nexus report` | Generate full project status report | | `nexus standup` | Generate morning standup | | `nexus eod` | Generate end-of-day summary | | `nexus risk` | Show risk register and active blockers | | `nexus decisions` | List all pending escalated decisions | | `nexus memory` | Show cross-session context summary | | `nexus log` | Show last 20 log entries | | `nexus map` | Show session routing table | | `nexus add task` | Manually add a task to the backlog | | `nexus pause` | Stop monitoring, save full state | | `nexus resume` | Resume from saved state | --- ## Drive Paths NEXUS sets paths dynamically during the interview — no hardcoded locations. ``` $PROJECT = [your project's Google Drive root — set in Q1.3] $DOCS = [your project's docs folder — set in Q1.4] $ORCH = $PROJECT/orchestrator/ ``` NEXUS reads your project's CLAUDE.md and README.md at startup and follows your project's folder rules for all file placement. Documentation files (backlog, sprint board, decisions log) go in $DOCS. Routing mechanics (inbox, outbox, cowork folders) go in $ORCH. --- ## Drive Structure (v2.0 Full) ``` $ORCH/ backlog.md ← master task list sprint-board.md ← current sprint project-plan.md ← WBS + dependencies decisions-log.md ← all decisions ever made risk-register.md ← risks and blockers routing-table.md ← session routing map status.md ← live NEXUS status README.md ← folder guide inbox/[chat-name]/ ← raw task drops from Chats outbox/[chat-name]/ ← results for Chats cowork/[name]/instructions/ ← work assigned to Coworks cowork/[name]/results/ ← completed work from Coworks decisions/ ← escalated, awaiting Amir decisions/resolved/ ← completed decisions log/nexus-[date].log ← daily audit log log/processed.log ← dedup list log/active-tasks.md ← live task tracker log/task-counter.txt ← incrementing task ID sessions/ ← session maps per run archive/ ← completed sprints + old tasks ```

View on GitHub

Back to search Back to installation

lovable-mcp-setup

Connect Claude Code (or another agent) to the Lovable MCP server so it can work on Lovable projects directly — reading files and diffs, checking analytics, and running SQL against a Lovable Cloud database via query_database. Use when setting up the Lovable connection for the first time, when a Lovable tool call fails with an auth or permission error, when /mcp shows lovable as disconnected, when deciding whether a Lovable action costs credits, or when you need the project_id for a Lovable project. Also use before any first use of query_database, since the safety rules live here.

2026-10-11

lovabledatabase

Do database work on a Lovable Cloud project directly, without going through Lovable's chat — schema, RLS policies, functions, triggers, scheduled jobs, backfills, diagnostics, and the bookkeeping that has to follow. Use this whenever work touches a Lovable app's database: adding or changing tables and columns, writing or auditing RLS policies, creating RPCs, triggers or cron jobs, investigating a query that returns nothing or everything, or deciding which channel a change belongs in. The Lovable MCP server's query_database runs SQL directly, so most work needs no Lovable chat at all — but it writes no migration file and regenerates no types, and that gap is what this skill exists to manage. Use it even when the request sounds like plain SQL ("add a status column", "why can users see each other's rows").

2026-10-11

marketing-expert

Use for creative, design, and marketing work spanning brand identity (logos, naming, color, typography, brand systems), digital and product design (websites, landing pages, app UI, dashboards, design systems, UX critique), copywriting (headlines, taglines, hero copy, ad copy, microcopy, voice and tone), marketing strategy (positioning, audience, messaging, GTM, campaigns), and creative rights / IP (copyright, trademark, fair use, licensing, AI authorship, asset clearance). Triggers on: design / critique / improve a logo, brand, name, site, page, screen, or design system; write / rewrite copy; position or differentiate a product; "is this fair use," "can I trademark / use / own this," "do I own this AI-generated work," "is this name taken"; compound briefs ("name + logo + landing page + pitch"); vague briefs ("make this more modern / on-brand / professional"). Surface proactively when work involves real assets, brand-adjacent names, AI-generated commercial output, or competitor references.

2026-10-11

View all from this repository

nexus

Source facts

Install options

Install with an AI assistant

Review the source files

More from this repository

Install with an AI assistant

More from this repository