Ops Gtm

OPS ► GTM COMMAND CENTER

Strategy layer that sits on top of /marketing. /marketing runs campaigns; /gtm decides what to run, across paid, unpaid, sales, and AI-automation channels, and then hands the executable pieces to /marketing via the Skill tool.

Runtime Context

Before executing, load available context:

1. Preferences: Read ${CLAUDE_PLUGIN_DATA_DIR:-$HOME/.claude/plugins/data/ops-ops-marketplace}/preferences.json

   • timezone — timestamp all output correctly
   • gtm_default_project, gtm_default_audience, gtm_brand_voice — project-level defaults when set
   • gtm_monthly_budget, gtm_stage (pre-launch|beta|ga|scale) — used to tune channel recommendations

2. Cached plans: List ${CLAUDE_PLUGIN_DATA_DIR}/gtm/*.md to surface recent plans. Never overwrite a prior plan file — always append a new dated file.

3. Daemon health: Read ${CLAUDE_PLUGIN_DATA_DIR}/daemon-health.json. If action_needed is not null, surface it before running any long planning flow.

4. Repo auto-scan (background): For the current working directory, in parallel and with run_in_background: true:

   • git remote -v → infer project slug and org
   • cat README.md (or README.*) → product description, ICP hints
   • Glob package.json, pyproject.toml, Cargo.toml, go.mod → tech stack
   • Glob .planning/**/*.md and docs/**/*.md → prior briefs, positioning notes
   • Resolve: project name, one-line pitch, primary tech, existing channels hinted in the repo

5. /marketing credential probe (read-only): Do NOT re-resolve API keys in this skill. Instead, when the user asks to launch something, delegate to /marketing via the Skill tool — /marketing owns the credential resolution chain (userConfig → env vars → Doppler → Dashlane → Keychain → gcloud ADC).

Agent Teams support

If CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is set, use Agent Teams for the plan sub-command so the four research agents share context and report progress in real-time:

TeamCreate("gtm-research-team")
Agent(team_name="gtm-research-team", name="paid-research",       prompt="Research paid acquisition channels that fit ${PROJECT_TYPE} at ${STAGE} with $${MONTHLY_BUDGET}/mo. Return a ranked list with fit signals, expected CAC, and which /marketing sub-command (if any) executes each channel.")
Agent(team_name="gtm-research-team", name="unpaid-research",     prompt="Research organic channels (SEO, content, community, PR, partnerships, referrals, lifecycle email) for ${PROJECT_TYPE} at ${STAGE}. Return ranked list with effort estimate, time-to-signal, and /marketing sub-command mapping.")
Agent(team_name="gtm-research-team", name="sales-research",      prompt="Design a sales motion for ${PROJECT_TYPE}: outbound vs inbound vs PLG vs channel. Propose a 30/60/90 plan with concrete activities and target metrics.")
Agent(team_name="gtm-research-team", name="automation-research", prompt="Propose AI-automation recipes: lead enrichment, personalized outreach, content ops, lifecycle copy, support deflection, lead scoring. For each, list trigger, model/tool stack, and where it plugs into /marketing.")

If the flag is NOT set, use standard fire-and-forget subagents with the same four prompts.

Sub-command Routing

Route $ARGUMENTS to the correct section below:

Arguments are free-form — treat /gtm plan for my-project $2k/mo pre-launch as equivalent to plan with intake values pre-filled.

Project Intake

Before producing any plan section, make sure the following are known. Source them in this order — only ask the user for what's still missing.

1. Auto-scan the repo (step 4 of Runtime Context).
2. Read prefs for gtm_default_* keys.
3. Parse free-text in $ARGUMENTS for budget, stage, and project name.
4. AskUserQuestion for gaps — respect Rule 1 (max 4 options per call; batch if needed).

Use these four questions in order, skipping any already known:

• Project type — [B2B SaaS, B2C product, Marketplace, Dev tool / API]
• Stage — [Pre-launch, Beta / early access, GA (live), Scale (growth)]
• Primary goal (next 90 days) — [Awareness, Signups / leads, Revenue, Retention / expansion]
• Monthly budget tier — [<$1k, $1k–5k, $5k–25k, $25k+]

If a Rule-1 batch overflows (e.g. more than 4 project types needed), paginate with [More options...] as the 4th slot.

Per Rule 3, never silently skip an intake question — if the user hits Escape, offer [Paste manually] / [Use default] / [Skip this only].

Channel / Avenue Catalog

This is the source-of-truth list the planner draws from. Each row: avenue → fit signals → cost profile → execution path. The execution path is what makes /gtm seamless with /marketing: if a /marketing sub-command exists for a channel, the plan recommends it by name; otherwise the channel is marked manual and the plan includes templated next-actions instead.

Paid

Unpaid (Organic)

Sales

AI Automation

plan — full GTM plan (default)

1. Intake per Project Intake section.

2. Spawn the four research agents (Agent Teams if flag set, else fire-and-forget).

3. Assemble the plan in this order:

   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    GTM PLAN — [project]  ([stage], [goal], $[budget]/mo)
    [timestamp]
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    POSITIONING
     ICP:            [one sentence]
     Value prop:     [one sentence]
     Messaging:      [3 pillars, comma-separated]
   
    PAID (next 30 days)
     1. [Channel]  $[X]/mo  Expected CAC: $[X]  Execute: /marketing ads
     2. [Channel]  $[X]/mo  Expected CAC: $[X]  Execute: manual
   
    UNPAID (next 90 days)
     1. [Channel]  Effort: [L/M/H]  Signal in: [X wks]  Execute: /marketing seo
     2. [Channel]  ...
   
    SALES
     Motion:         [Outbound / Inbound / PLG / Channel]
     30/60/90:       [one line summary]
   
    AUTOMATION (AI-powered)
     1. [Recipe]     Plugs into: /marketing email
     2. [Recipe]     Plugs into: outbound
   
    KPIs
     North star:     [metric + target]
     Leading:        [3 leading indicators]
   
    BUDGET ROLLUP
     Paid:           $[X]/mo
     Tools/SaaS:     $[X]/mo
     Total:          $[X]/mo
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
   

4. Persist the plan to ${CLAUDE_PLUGIN_DATA_DIR}/gtm/<project-slug>-$(date +%Y-%m-%d).md (create the directory if missing). Never overwrite — if the file exists, append -v2, -v3.

5. Handoff prompt — end with an AskUserQuestion (≤4 options per Rule 1):

   • [Launch via /marketing] — hand top executable items to /marketing
   • [Save only] — keep the plan file, take no action
   • [Refine plan] — re-run with adjusted intake
   • [Schedule follow-up] — revisit in 7/30 days (delegate to /ops cron if available)

On [Launch via /marketing], iterate through plan items whose execution path starts with /marketing and invoke them one at a time via the Skill tool:

Skill("ops-marketing", args="campaigns")
Skill("ops-marketing", args="ads")
Skill("ops-marketing", args="email")

Never silently skip an item (Rule 3) — for each, ask [Launch, Skip, Edit first].

Per Rule 5, do NOT auto-launch anything that spends money or sends to real recipients without explicit per-item confirmation. The plan recommends; /marketing executes; the user approves each paid or outbound action.

paid — paid acquisition deep-dive

Same intake, but only spawn the paid-research agent. Render just the PAID block, with up to 5 channel rows and for each:

• Expected CAC range (cite source or reasoning)
• Creative angle suggestion
• Starting budget
• Attribution setup note (how /marketing attribution will measure it)
• Execute: line pointing to the /marketing sub-command or manual

End with the same handoff prompt as plan, scoped to paid channels.

unpaid — organic deep-dive

Spawn unpaid-research only. For each recommended channel output:

• Why it fits (fit signals from intake)
• First-30-days concrete actions (3 bullets)
• Time-to-signal
• Measurement: which /marketing sub-command tracks it (/marketing seo, /marketing email, /marketing instagram) or manual + tool
• Lifecycle email is first-class here — recommend specific Klaviyo flows and note that /marketing email can scaffold them

sales — sales motion

Spawn sales-research only. Produce:

1. Motion pick with one-paragraph rationale (Outbound / Inbound / PLG / Channel — Rule 1: these are the exact 4 options if you need the user to choose).
2. ICP slice — which segment to hit first, sized from intake.
3. 30/60/90 plan — week-by-week activities (e.g. week 1: build list of 500 accounts; week 2: 200 outbound opens; …).
4. Tooling — CRM, enrichment, sequencer; note when AI automation from the automation section fits.
5. Handoff — offer [Automate outreach via /gtm automation, Skip, Edit motion].

automation — AI automation playbook

Spawn automation-research only. Render each recipe with:

• Trigger condition (what event fires it)
• Model / tool stack (default to Claude + the ops tool listed in the catalog)
• Integration point — specifically, which /marketing sub-command or external system it reads from and writes to
• Risk flags (PII, cost, brand-voice drift) and the guardrails that mitigate them
• Starter prompt / pseudo-code snippet so the user can copy-paste

End with AskUserQuestion offering [Scaffold recipe now, Save only, Pick different recipe].

launch — 30/60/90 launch calendar

Used at stage = pre-launch or beta. Produces a week-by-week calendar of launch activities across all four avenues.

Required pre-flight checklist (render and check what's ready vs open):

☐ Positioning brief  →  /gtm brief
☐ Landing page live  →  manual
☐ Analytics wired    →  /marketing setup (GA4 + GSC)
☐ Email capture live →  /marketing email  (Klaviyo list)
☐ Ad accounts ready  →  /marketing setup (Meta, Google)
☐ Social handles claimed   →  manual
☐ Product Hunt / HN plan   →  this skill, below
☐ Press / creator list     →  manual

Launch-day playbook: render hourly timeline for T-7d through T+14d, with concrete actions and the /marketing command (or manual step) for each.

brief — one-page positioning brief

Fast path — does NOT spawn the four research agents. Just intake + a single call to write the brief. Fields: ICP, Pain, Value prop, 3 messaging pillars, Proof points, Anti-positioning (who it's NOT for). Save to ${CLAUDE_PLUGIN_DATA_DIR}/gtm/<project-slug>-brief-$(date +%Y-%m-%d).md.

This is the cheapest entry point — recommend running /gtm brief before /gtm plan when the project's positioning isn't already written down.

setup

Configure GTM defaults so subsequent runs skip the intake questions. Per Rule 4, run every Bash call with run_in_background: true unless the result is needed for the very next decision.

Auto-scan (background) first:

# Existing prefs
jq -r '{project: .gtm_default_project, audience: .gtm_default_audience, voice: .gtm_brand_voice, budget: .gtm_monthly_budget, stage: .gtm_stage}' "$PREFS_PATH" 2>/dev/null

# Repo signals
git -C "$PWD" remote get-url origin 2>/dev/null
head -50 README.md 2>/dev/null
ls .planning/ 2>/dev/null

Then prompt for what's missing (≤4 options per Rule 1), in this order:

1. Default project name — free-text, pre-filled from git remote slug
2. Project type — [B2B SaaS, B2C product, Marketplace, Dev tool / API]
3. Stage — [Pre-launch, Beta, GA, Scale]
4. Monthly budget tier — [<$1k, $1k–5k, $5k–25k, $25k+]
5. Brand voice — [Playful, Professional, Technical, Bold]

Save to $PREFS_PATH as gtm_default_project, gtm_project_type, gtm_stage, gtm_monthly_budget, gtm_brand_voice. Never write API keys here — GTM does not own any credentials (/marketing does).

Finish with a smoke test: run /gtm brief in background and report ✓ setup complete — try /gtm plan or ✗ [error].

Plugin Rules compliance (required reading)

All behavior above respects claude-ops/CLAUDE.md:

• Rule 0 — public repo: every example above uses your-project, you@example.com, <YOUR_TOKEN>. Never save user-specific data outside $PREFS_PATH or ${CLAUDE_PLUGIN_DATA_DIR}/gtm/.
• Rule 1 — ≤4 options per AskUserQuestion: every prompt in this file lists exactly ≤4. Paginate with [More options...] if a dynamic list grows past 4.
• Rule 3 — never auto-skip: on launch handoff, every plan item gets an explicit [Launch, Skip, Edit first] prompt. No silent skipping.
• Rule 4 — background by default during setup: all Bash calls in the setup flow use run_in_background: true.
• Rule 5 — destructive actions need explicit confirmation: /gtm never spends money or sends messages directly — it delegates to /marketing, which owns per-action confirmation. Recommendations are just that.

Why this skill is thin

/gtm deliberately does NOT re-implement channel APIs. Credential resolution, curl calls, daemon data, and dashboards all live in /marketing. This skill is a strategy-and-handoff layer. If a capability belongs to a channel API, add it to /marketing; if it belongs to planning, it goes here.

Source: davepoon/buildwithclaude → plugins/claude-ops/skills/ops-gtm/SKILL.md