Ga4 Monitor logo

Ga4 Monitor

Organization
cognyai
ga4-monitor

Weekly GA4 check — sessions, users, key-event conversions, revenue, channel mix — with week-over-week deltas and tracking-break alerts

Overview

Publishercognyai
Repositoryclaude-code-marketing-skills
Skill namega4-monitor
Stars
111
Forks
15
Bundled files
Instructions only
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by cognyai on GitHub. Read the source before you install it.

Installation

Install the Ga4 Monitor AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/cognyai/claude-code-marketing-skills/tree/main/skills/ga4-monitor
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/cognyai/claude-code-marketing-skills.git /tmp/claude-code-marketing-skills
mkdir -p .claude/skills
cp -r /tmp/claude-code-marketing-skills/skills/ga4-monitor .claude/skills/ga4-monitor
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Ga4 Monitor in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Ga4 Monitor on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Ga4 Monitor is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

GA4 Monitor

A recurring weekly check on a GA4 property. It confirms data is still flowing, reports traffic and conversions week-over-week, and — most importantly — catches the silent failure mode: a key event that stopped firing because a tag broke.

Requires: Cogny Agent subscription ($9/mo) — Sign up

Designed to be scheduled weekly via /loop or /schedule.

Works with

JobGA4 (primary)PlausibleFathom AnalyticsSimple AnalyticsAdobe Analytics
Heartbeattool_run_realtime_reportrealtime_visitorscurrent_visitorsstats (start/end = today)run_report on yesterday
Traffic WoWtool_run_reportaggregateaggregate (entity: pageview)statsrun_report
Conversions WoWtool_list_conversion_events + tool_run_reportbreakdown (event:goal)list_events + aggregate (entity: event)eventsrun_report (your success-event metrics)
Channel mixtool_run_report (sessionDefaultChannelGroup)breakdown (visit:source, visit:utm_campaign)aggregate with field_grouping: referrer_hostnamereferrersrun_report (marketing-channel dimension)

Fathom Analytics here is the web analytics product, not the Fathom call recorder. Enhanced when connected, with any of them: Shopify / WooCommerce orders and PostHog / Mixpanel sign-ups confirm whether a conversion drop is real or a tracking break (Step 4). Several connected: run the checks on each and show them side by side, never summed.

Prerequisites Check

If no mcp__cogny__* tools exist (no list_integrations, no cogny_status), print this and stop:

GA4 Monitor needs your marketing data through Cogny's MCP gateway.

1. Sign up at https://cogny.com and connect Google Analytics 4 (or Plausible, Fathom
   Analytics, Simple Analytics or Adobe Analytics).
2. Add Cogny to .mcp.json in your project:
   {"mcpServers": {"cogny": {"type": "http", "url": "https://app.cogny.com/mcp"}}}
3. Restart Claude Code; it opens a browser for the OAuth login on first use.

Nothing to run without Cogny.

Otherwise call list_integrations once. Use every healthy web-analytics provider from the table above; name any broken one. health_status misses revoked tokens, so make one cheap read per provider (GA4 tool_list_account_summaries, Adobe list_report_suites, Fathom list_sites, …): a 401, Forbidden or scope error is 🔴 "reconnect ", with no retry. If a tool named here is missing, call refresh_mcp_discovery once; new tools only load after the client reconnects, so skip that check this run and mark it ⚠ "needs reconnect". If no web-analytics provider is connected, say "connect Google Analytics 4 (or Plausible, Fathom, Simple Analytics, Adobe) in Cogny" and stop.

Usage

/ga4-monitor — weekly health check of the connected GA4 property.

For a full one-time review, run /ga4-audit. This is the light weekly rhythm.

Steps

1. Read the prior snapshot

Call read_context_node on insights/ga4/monitor/latest. If present, it holds last week's numbers — report every metric as a delta against it. If absent, this is the first run: report absolutes and note deltas begin next week.

Tool blocks below are GA4. For other providers use the "Works with" row; what differs is noted in each step.

2. Data-flow heartbeat

tool_run_realtime_report   (property_id, metrics: activeUsers)

If realtime shows zero active users on a site with normal traffic, tagging may be down. Note it, but confirm against the 7-day report before alarming.

Other providers: Plausible__realtime_visitors (site_id), Fathom_Analytics__current_visitors (site_id). Simple and Adobe have no realtime tool, so use today's (Simple stats) or yesterday's (Adobe run_report) totals instead.

3. Traffic & engagement (last 7d vs prior 7d)

tool_run_report   (property_id,
  date_ranges: [{"startDate": "7daysAgo", "endDate": "yesterday"},
                {"startDate": "14daysAgo", "endDate": "8daysAgo"}],
  metrics: ["sessions", "totalUsers", "newUsers", "engagementRate", "averageSessionDuration"])

Pin the windows exactly like this: both are 7 full days and today's partial day is excluded, so the delta isn't skewed. Use the same two ranges in steps 4 and 5, and the same calendar dates for other providers. Report each metric with its WoW delta.

ProviderCall (run twice: this week, prior week)Metric names
Plausibleaggregate (site_id, date_range: [start, end])visitors, visits, bounce_rate, visit_duration
Fathomaggregate (entity: pageview, entity_id: <site id>, date_from, date_to)visits, uniques, bounce_rate, avg_duration
Simple Analyticsstats (hostname, start, end)pageviews, visitors
Adoberun_report (rsid, start_date, end_date, dimension: variables/daterangeday, metrics)metrics/visits, metrics/visitors

Map "sessions" to visits and "users" to visitors. These tools have no engagement rate; report bounce rate instead and say so.

4. Key-event conversions (the critical check)

tool_list_conversion_events   (property_id)
tool_run_report   (property_id, same two date_ranges as step 3,
  dimensions: eventName,
  metrics: eventCount, keyEvents, totalRevenue)

Use keyEvents, not the deprecated conversions metric. Join the report against tool_list_conversion_events, not the other way round: a key event with zero events in both weeks returns no row at all, so start from the key-event list and fill missing rows with 0. A key event at 0 / 0 is 🟡 "never fires" (a dead or misnamed event; event names are case-sensitive, so check /gtm-audit for a first_login vs FirstLogin mismatch).

🔴 Alert if any key event recorded conversions last week and zero this week — the tag or GTM trigger has almost certainly broken. This is the single most important check in the skill.

🟡 Flag key events that dropped >50% WoW without an obvious traffic cause.

Same check on other providers, run on each conversion:

ProviderConversions this week vs lastCan't check
Plausiblebreakdown (dimension: event:goal, metrics: [visitors, events, conversion_rate]) per weekWhich goals count as conversions (all goals do); revenue unless revenue goals exist
Fathomlist_events (site_id), then aggregate (entity: event, entity_id: <event id>) per eventRevenue unless the event carries a value
Simple Analyticsevents (hostname, start, end)Key-event config; revenue
Adoberun_report with your success-event metrics (find them with list_metrics)Which events are "key" lives in the report suite, not the API

Confirm a drop against the source of truth (enhanced when connected). Before calling a drop a tracking break, check the same two weeks where the conversion really happens: Shopify tool_list_orders (query_filter: "created_at:>=YYYY-MM-DD") or WooCommerce get_sales_report (date_min, date_max) for orders; PostHog list_insights (project_id, search: "signup") → get_insight, or Mixpanel run_segmentation (event, from_date, to_date) for sign-ups. Truth flat + analytics down = 🔴 tracking break. Both down = a real demand drop; say so and skip the tracking alarm.

Only GA4 exposes a key-event config (tool_list_conversion_events). For the others, say in the output that "key event" means the goals or events the user named.

5. Channel mix

tool_run_report   (property_id, same two date_ranges as step 3,
  dimensions: sessionDefaultChannelGroup,
  metrics: sessions, keyEvents)

🟡 Alert on:

  • A jump in Unassigned / (not set) sessions — UTMs broke somewhere
  • A Direct spike — often a tagging or redirect problem
  • A channel that fell off a cliff (campaign ended, or tracking lost)

Other providers: Plausible breakdown (visit:source, and visit:utm_campaign for the UTM check), Fathom aggregate (field_grouping: referrer_hostname), Simple referrers, Adobe run_report on your marketing-channel dimension (list_dimensions if unsure). These group by source rather than GA4's channel groups, so the Unassigned check becomes "a jump in direct / no-referrer visits".

6. Output

GA4 Monitor — [Property Name] ([Provider])
Week of [YYYY-MM-DD] – [YYYY-MM-DD]

⚠️  Needs attention this week
- [Critical/important first, or "Nothing — data is flowing clean."]

Traffic (7d vs prior 7d)
- Sessions ........ XX,XXX  (+/-X% WoW)
- Users ........... XX,XXX  (+/-X% WoW)
- Engagement rate . XX%     (+/-X pts WoW)

Conversions (7d vs prior 7d)
| Key event | This week | Last week | WoW | Revenue |
|-----------|-----------|-----------|-----|---------|

Channel mix
| Channel | Sessions (WoW) | Conv (WoW) |
|---------|----------------|------------|

Recommended actions
1. ...

7. Save this week's snapshot

Call write_context_node on insights/ga4/monitor/latest (other providers: insights/<provider>/monitor/latest, e.g. insights/plausible/monitor/latest) with this week's metrics (sessions, users, engagement, per-key-event conversions + revenue, channel mix) so next week can compute deltas. Also write a dated copy to insights/ga4/monitor/<YYYY-MM-DD>.

8. Record findings — selectively

Only file a create_finding for genuine problems — a key event that stopped, a channel that collapsed, a (not set) spike. A clean week files nothing.

json
{
  "title": "GA4 key event 'generate_lead' dropped to 0 this week",
  "body": "'generate_lead' recorded 96 conversions last week and 0 this week while sessions held steady. The form tag or its GTM trigger has broken. Check GTM Preview on the form page and the GA4 DebugView.",
  "action_type": "tracking_break",
  "expected_outcome": "Lead tracking restored; Ads bidding gets signal again",
  "estimated_impact_usd": 0,
  "priority": "critical"
}

Action types: tracking_break, data_quality, conversion_setup, channel_anomaly.

Recording findings

Use create_ticket when the workspace has it (Cloud) and create_finding otherwise (Solo). The fields differ; see section 2 of ../marketing-engineer/references/mcp-routing.md.

Critical rules

  1. Trends over snapshots. Every metric ships with a WoW delta.
  2. A key event at zero is critical — it usually means a broken tag, and it corrupts Ads bidding too. Cross-check /gtm-monitor if GTM is connected.
  3. Lead with what's broken. A clean week is one short paragraph.
  4. Read-only. Surface issues; never change the property.
  5. Name the provider on every number. Plausible visitors, Fathom uniques and GA4 users are counted differently; never add them together.

Frequently asked questions

What does the Ga4 Monitor AI skill do?

Weekly GA4 check — sessions, users, key-event conversions, revenue, channel mix — with week-over-week deltas and tracking-break alerts

Why use Ga4 Monitor on TypingMind?

Because you install it once and use it with any model. Ga4 Monitor is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Ga4 Monitor in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/cognyai/claude-code-marketing-skills/tree/main/skills/ga4-monitor. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Ga4 Monitor?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Ga4 Monitor?

As many as you like. As long as a model supports skills, you can use Ga4 Monitor with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Ga4 Monitor AI skill free?

It is published on GitHub by cognyai. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇