Skip to main content

ce-product-pulse

A skill from Every's compound-engineering plugin that generates a time-windowed (24h / 7d / custom) one-page product-health report — engagement, value realization, latency and error rates, plus signals worth investigating. It reads `STRATEGY.md` and `.compound-engineering/config.local.yaml`, pulls from analytics, tracing, and optional databases, and writes a timestamped report to `docs/pulse-reports/` while keeping no PII in saved output.

Get a fast 'how is the product actually doing' snapshot tied to the metrics declared in STRATEGY.md

Source Every
License MIT
First documented
Receipts TODO

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

  • show me the pulse
  • weekly product recap
  • how are we doing this week

What it does

ce-product-pulse generates a compact, single-page report on product health across a time window — typically 24h, 7d, or a custom duration. See skills/ce-product-pulse in Every’s compound-engineering plugin. You run it when you want a quick snapshot — “show me the pulse,” “weekly recap,” “how are we doing?” The report covers engagement metrics, value realization, system performance (latency and error rates), recommended follow-ups, and signals worth investigating.

It reads STRATEGY.md if present (so the report is framed against the product’s real key metrics rather than generic ones) and .compound-engineering/config.local.yaml for configuration, then pulls from external data sources — analytics, tracing, and optionally databases — and writes a timestamped report to docs/pulse-reports/YYYY-MM-DD_HH-MM.md. It keeps no PII in the saved output. In the plugin’s loop, it closes what ce-strategy opens: strategy declares the metrics that matter, and product-pulse measures the product against them while surfacing what to look at next.

When to use it

  • A weekly or daily product health check — usage, errors, and latency in one page
  • After a launch — did the numbers move, what broke, what to investigate
  • Grounding a planning or ideation session in what users actually experienced
  • Reporting up without hand-assembling a dashboard screenshot

When not to reach for it:

  • Deep incident forensics — this is a pulse, not a root-cause investigation (pair it with ce-debug)
  • Marketing / funnel analytics (GA4, attribution) — that’s a different lane (see analytics-tracking)
  • Anywhere you’d need PII in the artifact — it deliberately strips it

Install

Part of the compound-engineering plugin (not a standalone ~/.claude/skills/ drop-in). In Claude Code:

/plugin marketplace add EveryInc/compound-engineering-plugin
/plugin install compound-engineering

Then invoke /ce-product-pulse (e.g. /ce-product-pulse 7d). The plugin namespaces every skill with ce-. MIT-licensed.

What a session looks like

  1. /ce-product-pulse 7d on a Monday morning.
  2. It reads STRATEGY.md to know which metrics count, and .compound-engineering/config.local.yaml for data-source configuration.
  3. It pulls data: engagement and value metrics from analytics, latency and error rates from tracing, optionally a database query.
  4. It assembles a one-page report: what users experienced, how the product performed, signals worth investigating, recommended follow-ups.
  5. It writes docs/pulse-reports/2026-06-01_09-00.md — timestamped and PII-free.
  6. The “signals worth investigating” section becomes input to the next ce-plan or ce-debug.

The discipline that makes it work: the report is framed against STRATEGY.md, not generic vanity metrics. The pulse only means something if the numbers it reports are the ones the strategy said mattered — which is why the two skills are designed as bookends of the same loop.

Receipts

TODO — to be filled in from a real session. Once product-pulse has been wired to real data sources, this section will capture: which sources it connected to automatically vs. needed manual config in .compound-engineering/config.local.yaml, whether framing the report against STRATEGY.md’s metrics produced a more useful pulse than a generic dashboard, how clean the PII-stripping actually was, and whether the “signals worth investigating” section pointed at real issues or noise.

Source and attribution

From Every’s compound-engineering plugin — the MIT-licensed plugin (Dan Shipper & Kieran Klaassen) packaging Every’s engineering workflow.

License: MIT.

Quoting the skill’s own framing verbatim: a report on “what users experienced and how the product performed — usage, quality, errors, signals worth investigating.” It is the downstream measurement half of the ce-strategyce-product-pulse loop.