╭──────────────────────────────────────────────────────────────────────────────╮
│  /ops:go  ►  MORNING BRIEFING                              2026-04-12  09:03 │
├─────────────────────────────────┬────────────────────────────────────────────┤
│  INFRA    ████████████████  ok  │  ECS: 4/4 healthy  RDS: ok  Redis: ok     │
│  CI/CD    ████████████░░░░  75% │  3 passing  1 failing  (my-api #847)  │
│  INBOX    ░░░░░░░░░░░░░░░░  14  │  Slack: 9  Telegram: 3  Gmail: 2 unread   │
│  PRs      ████████████████  3   │  3 ready to merge  1 needs review          │
│  SPRINT   ████████████░░░░  67% │  Sprint 24  —  8 of 12 issues complete     │
│  REVENUE  ████████████████  $   │  $2,847 MTD  ↑12% vs last month           │
├─────────────────────────────────┴────────────────────────────────────────────┤
│  Next action: merge feat/user-profile  ·  fix my-api CI  ·  reply @alice    │
╰──────────────────────────────────────────────────────────────────────────────╯

Turn Claude Code into a complete business operating system — infrastructure health, CI/CD status, unified inbox, open PRs, sprint state, revenue snapshot (Stripe + RevenueCat + AWS), and autonomous C-suite agents that act on your behalf.

---

What's new in v3.6

Current: v3.6.1. One plugin, three harnesses, official skill shape.

Hermes	Installer mirrors skills into ~/.hermes/skills for working slash commands and links hermes-plugin/ for skill_view("ops:*").
Grok	Loads the Claude plugin as-is. No .grok-plugin.
Skills	Third-person trigger descriptions, shared preamble, ops-rules (plugin-root CLAUDE.md is a pointer). Oversized skills split into references/.
Versions	plugin.json = Hermes plugin.yaml = installer pin. ops-release keeps them together.
MCP	Plugin .mcp.json is empty. Servers start on demand, not once per session.

Ports: claude-ops/docs/harness-ports.md. Rules: skills/ops-rules/SKILL.md. Full notes: CHANGELOG.

---

What's new in v2.0

v2 turns claude-ops from a briefing + comms surface into an autonomy layer for Claude Code itself. Purely additive — no v1 behaviour changes by default. See claude-ops/CHANGELOG.md and docs/migrating-from-v1.md.

Capability	Skill	Doc
Post-merge + build-failure auto-fix (PostToolUse hooks → headless Haiku fixer)	/ops:deploy-fix	deploy-fix.md
Pre-installed specialist agents + silent general-purpose → specialist routing	(transparent)	agents.md
Universal safety hooks: secret-scan, rm -rf anchor block, main push warn	(always-on)	safety-hooks.md
Recap marquee — multi-session digest in tmux status-right / statusLine	/ops:recap	recap.md
Multi-account Claude Max rotator with launchd daemon + AI-brain (jittered post-rotation respawn + /login re-auth so a freshly-rotated account isn't hit by a synchronized first-call burst)	/ops:rotate, /ops:rotate-setup	CHANGELOG
Periodic Task* tracking nudge	(PostToolUse hook)	CHANGELOG
Windsor.ai live marketing data source for ops dashboards, marketing, socials & ecom	/ops:marketing	windsor-ai.md

Quick start for the auto-fix subsystem

# 1. Upgrade
/plugin update ops@ops-marketplace

# 2. Run the wizard (hits new steps 6.5a–6.5d for v2 toggles)
/ops:setup

# 3. Map your repos to their deploy URLs
/ops:deploy-fix configure
# (opens ~/.claude/config/post-merge-services.json)

# 4. From now on, every `gh pr merge` you run from Claude Code will:
#    - poll the deploy workflow
#    - curl /health on success
#    - verify /version returns the merged SHA
#    - on failure: auto-rerun transients, OR dispatch a Haiku deploy-fixer
/ops:deploy-fix          # see status / budget / live runs

Per-repo budget caps (default 3/hour), single-flight locks, and content-hash dedup prevent runaway spending. Notifications route via macos/ntfy/pushover/discord/telegram/none. Every toggle is spacebar-toggleable in /plugins settings.

---

Quick Start

# 1. Add the marketplace (ships ops + desktop-act companion)
/plugin marketplace add Lifecycle-Innovations-Limited/claude-ops

# 2. Install the ops plugin
/plugin install ops@ops-marketplace
# Optional explicit companion (also run by /ops:setup and /ops:update):
# /plugin install desktop-act@ops-marketplace

# 3. Run the guided setup wizard (co-installs desktop-act by default)
/ops:setup

Cross-CLI install (Claude Code + Codex + Gemini + OpenClaw + Hermes + OpenCode)

git clone https://github.com/Lifecycle-Innovations-Limited/claude-ops.git
cd claude-ops/installer
npm install
node bin/claude-ops-installer.mjs install

One command mirrors upstream skills + binstubs into every detected CLI's expected layout from a single central config (~/.config/claude-ops-installer/config.yaml). See installer/README.md for the schema, supported agents, and verify / doctor / update / uninstall subcommands.

[!TIP] The wizard installs the background daemon EARLY (Step 2c). While you're still answering "connect Slack? [OAuth/Skip]" questions, briefing-pre-warm is already running every 2 minutes — pre-fetching ECS health, git state, PRs, CI, and unread counts. By the time setup finishes, your first /ops:go briefing loads in <3 seconds from warm cache instead of ~30s cold.

Local development:

git clone https://github.com/Lifecycle-Innovations-Limited/claude-ops.git
claude --plugin-dir ./claude-ops/claude-ops

---

Commands

All 66 skills, grouped by category:

🧭 Navigation	📊 Daily Ops
/ops — pixel-art dashboard	/ops:go — morning briefing
/ops:dash — same + hotkeys	/ops:next — priority next action
/ops:setup — guided wizard	/ops:inbox — deep-context inbox zero
/ops:uninstall — clean removal	/ops:comms — send/read any channel
	/ops:merge — autonomous PR pipeline
	/ops:relations — relationship manager
	/ops:vip — VIP list, answered first

🛠️ Project & Eng	💰 Business
/ops:projects — portfolio dashboard	/ops:revenue — Stripe + RevenueCat + AWS
/ops:linear — sprint board	/ops:ecom — Shopify operations
/ops:triage — cross-platform issues	/ops:marketing — Klaviyo/Meta/GA4/GSC
/ops:fires — incidents + all AWS	/ops:gtm — cross-channel GTM planner
/ops:deploy — ECS/Vercel/Actions	/ops:voice — native + Twilio + Bland + Zoom + smart join
/ops:monitor — Datadog/New Relic/OTEL	/ops:package — carrier-agnostic shipping
/ops:competitors — self-discovering competitor intel (v2.3)

🤖 Automation	🧰 Maintenance
/ops:orchestrate — parallel engine	/ops:speedup — GPU/ANE + power hogs + OS actions
/ops:yolo — 4 parallel C-suite agents	/ops:doctor — plugin auto-repair
/ops:integrate — add external service	/ops:daemon — launchd background brain
/ops:whatsapp-biz — catalog/orders	/ops:status — plugin + daemon health

Voice / phone / video (v2.9)

/ops:voice is a full voice surface — not just AI calls.

Subcommand	What it does
phone <number>	Native Phone.app dial via Continuity
facetime <handle> [--audio]	FaceTime video or audio
zoom start|join|schedule	Native Zoom + REST scheduling
join [--at now|next]	Auto-joins the current/next calendar meeting with smart AV defaults (cam/mic policy by attendee count, lid-state mic switching, Elgato auto-launch)
twilio-call, twilio-sms	Programmatic outbound voice + SMS
bland-call	AI agent phone call
tts, transcribe	ElevenLabs TTS + Groq Whisper

All outbound 1:1 channels (twilio, bland) go through the per-message approval gate (Rule 6).

Skill routing

flowchart TD
    U["User"] --> O["/ops"]
    O --> D["Dashboard"]
    D --> Daily["Daily Ops"]
    D --> Eng["Project & Eng"]
    D --> Biz["Business"]
    D --> Auto["Automation"]
    Daily --> go["/ops:go"]
    Daily --> inbox["/ops:inbox"]
    Daily --> merge["/ops:merge"]
    Eng --> projects["/ops:projects"]
    Eng --> linear["/ops:linear"]
    Eng --> fires["/ops:fires"]
    Biz --> revenue["/ops:revenue"]
    Biz --> ecom["/ops:ecom"]
    Biz --> marketing["/ops:marketing"]
    Auto --> yolo["/ops:yolo"]
    Auto --> orchestrate["/ops:orchestrate"]

---

Before / After

┌────────────────────────────────────────────┬──────────────────────────────────────────────┐
│  WITHOUT claude-ops                        │  WITH claude-ops                             │
├────────────────────────────────────────────┼──────────────────────────────────────────────┤
│  Open 6+ tabs every morning                │  /ops:go  ——  one command, done              │
│  Context-switch between Slack/Telegram/    │  /ops:inbox  ——  unified view, all channels  │
│  email                                     │                                              │
│  Manually review and merge PRs one by one  │  /ops:merge  ——  autonomous pipeline         │
│  SSH into servers to check health          │  /ops:fires  ——  terminal dashboard          │
│  Forget to track AWS spend                 │  /ops:revenue  ——  automatic cost snapshot   │
│  Switch between Linear and GitHub          │  /ops:linear + /ops:projects  ——  unified    │
└────────────────────────────────────────────┴──────────────────────────────────────────────┘

---

Integrations (22 services)

Most integrations offer two paths — MCP (zero-config OAuth) or CLI (fuller feature set). The setup wizard lets you choose per-integration.

SERVICE	MCP	CLI	WHAT YOU LOSE WITHOUT CLI
GitHub	—	gh (auto)	EVERYTHING — CI logs, PR merge, triage all require gh
AWS	—	aws (auto)	EVERYTHING — 17+ services probed by infra-monitor
Stripe	—	API key	Required for /ops:revenue MRR — web + desktop subs
RevenueCat	—	API key + project ID	Required for mobile-app subscription MRR
Linear	OAuth via Claude.ai (12 tools)	—	Nothing — fully covered
Vercel	OAuth via Claude.ai	—	Nothing — deploy status, build + runtime logs
Slack	OAuth via Claude.ai	local bot token	MCP covers most. Token adds: unlimited search, private ch
Gmail	OAuth (read)	gog (send+archive)	MCP = read-only. CLI = full autonomous inbox
Calendar	OAuth via Claude.ai	gog (read-only)	MCP has more features — either works
Sentry	OAuth via Claude.ai	sentry-cli	MCP covers triage. CLI adds source maps + releases
WhatsApp	—	wacli	EVERYTHING — no MCP exists
Telegram	—	bundled MCP server	EVERYTHING — plugin ships its own MTProto server
Shopify	—	Admin API + template	Store ops, order mgmt, inventory via /ops:ecom
Klaviyo	—	API key	Email/SMS campaigns via /ops:marketing
Meta Ads	—	API token	Paid-social reporting via /ops:marketing
GA4	—	service account	Analytics via /ops:marketing
GSC	—	service account	Search Console via /ops:marketing
Bland AI	—	API key	Outbound voice via /ops:voice
ElevenLabs	—	API key	TTS + cloning via /ops:voice
Whisper	—	API key	Transcription via /ops:voice
Twilio	—	API (Account SID + Auth Token)	Outbound voice + SMS via /ops:voice twilio-*
Zoom	Server-to-Server OAuth	API token	Native start/join (no creds) + REST schedule via /ops:voice zoom *
macOS Phone / FaceTime	—	URL schemes (tel:, facetime://, facetime-audio://)	Native via Continuity — no creds, no API
Elgato Camera Hub	—	auto-launched if installed	Virtual camera setup for meetings via /ops:voice join
GSD	—	auto-detected	Optional — roadmap state; degrades gracefully
Doppler	@dopplerhq/mcp-server (MCP)	doppler CLI (fallback)	Secrets manager; MCP server provides direct tool access

[!NOTE] infra-monitor now covers every AWS service you have IAM for — ECS, EC2, RDS, Lambda, S3, CloudFront, ALB/NLB, API Gateway, SQS, SNS, DynamoDB, ElastiCache, Route 53, ACM, CloudWatch, Budgets, IAM. Probes run in parallel; services you can't access are silently skipped.

---

Architecture

flowchart TB
    CC[Claude Code] --> S[Skills · 22]
    CC --> A[Agents · 12]
    CC --> H[Hooks]
    S & A & H --> RC[Runtime Context]
    RC --> P[preferences.json]
    RC --> M[memories/]
    RC --> SEC[Doppler · PW Mgr · Keychain · env]
    CC <--> D[ops-daemon<br/>launchd · 7 services]

All skills use pre-execution shell blocks (! fences) that gather data before model context loads — zero extra latency, minimal token overhead. The ops-daemon pre-warms briefing data so /ops:go hits warm cache.

Why the nested claude-ops/claude-ops/ directory? Claude Code's plugin marketplace system requires a two-level layout: the repo root acts as a marketplace container (with .claude-plugin/marketplace.json pointing "source": "./claude-ops"), while the inner directory is the actual plugin root (with .claude-plugin/plugin.json, skills, agents, etc.). This is how Claude Code resolves and caches plugins — it cannot be flattened.

claude-ops/                        ← marketplace root (this repo, this README)
├── .claude-plugin/
│   └── marketplace.json           # points to ./claude-ops as plugin source
├── README.md                      # ← you are here
│
└── claude-ops/                    ← plugin root (Claude Code loads from here)
    ├── .claude-plugin/plugin.json
    ├── CLAUDE.md                  # 5 non-negotiable plugin rules
    ├── skills/                    # 22 slash commands
    ├── agents/                    # 12 autonomous agents (Opus/Sonnet/Haiku)
    ├── bin/                       # ops-gather · ops-shopify-create · gog fallback
    ├── hooks/                     # SessionStart health check
    ├── telegram-server/           # bundled MCP server (gram.js)
    ├── templates/                 # Shopify Admin + app scaffolding
    ├── tests/                     # bash validation · test-no-secrets.sh
    └── .mcp.json                  # empty on purpose — MCP starts on demand, not per session

---

Agent Teams

Every ops skill that spawns agents supports Claude Code Agent Teams — a coordination layer where agents share context, report progress, and accept mid-flight steering.

Enable: Set CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 in your environment.

How it works: When the flag is set, skills create a named team and dispatch agents into it. Agents within a team can share findings (e.g., an inbox agent discovers a Slack message referencing an email thread, so the email agent prioritizes it) and you can steer priorities via SendMessage.

TeamCreate("fire-fixers")
Agent(team_name="fire-fixers", name="fix-ecs", ...)
Agent(team_name="fire-fixers", name="fix-ci", ...)
SendMessage(to="fix-ecs", content="This is P0, prioritize over CI")

Without the flag: Skills fall back