schematize
Skills · Claude Code · Agent Skills

Measure product with contracts and privacy — no negotiation.

An event is a contract and personal data is a liability: a versioned tracking plan, consent before collection, one definition per metric and experiments designed before they go live.

v0.2.0Disciplina · Analytics de produtogithub.com/schematizeme/skill-analytics

نگهداری‌شده توسط Lucas Andrade

schematize-analytics · v0.2.0bash
$ schematize install analytics
GitHub →دانلود مهارت (.zip)/analytics-help · github.com/schematizeme/schematize-cli

Non-negotiable floors

Never switched off "to go faster". Full list in the skill.

Events are contracts

Every event lives in the versioned tracking-plan.yml with an owner and closed properties; an event outside the plan fails CI.

object_action naming

Lowercase snake_case in the past tense (order_paid); no click, camelCase or event1.

No PII in events

Email, tax ID, phone, name, raw IP and tokens stay out; the user is the opaque internal ID.

Consent before collection

Analytics that identifies or cross-links requires granular opt-in, refusing as easy as accepting, and proof of consent.

First-party and minimization

Collect through your own domain, only what the declared purpose needs, with a defined retention.

One definition per metric

Formula, window, population, source and owner written down; two dashboards with the same name and different numbers is a data incident.

Experiment designed first

Hypothesis, primary metric, MDE, sample size and stopping rule written beforehand; peeking is banned and SRM fails the result.

Non-prd never reaches prd

Separate analytics project and key per environment; tests and CI use a disposable sink.

Commands

Installed in .claude/commands/. Run /analytics-help anytime.

/analytics-ccContext compact: writes the handoff to the archive and runs /compact.
/analytics-claudeCreates or merges the always-on CLAUDE.md for the product analytics baseline at the repo root (won't overwrite other skills' blocks).
/analytics-consentDesigns analytics consent: legal basis per purpose, opt-in, cookieless mode, proof and revocation.
/analytics-experimentDesigns an A/B experiment before it goes live: hypothesis, metric, MDE, sample size, guardrails, SRM and stopping rule.
/analytics-handoffWrites the context handoff (context.md + checklist.md) to the archive, without compacting.
/analytics-helpLists every command and what it does.
/analytics-loadForce-loads the whole product analytics normative body and applies it for the session.
/analytics-metricDefines a metric with a single definition (formula, window, population, source, owner) and the North Star.
/analytics-planCreates or evolves tracking-plan.yml (object_action events, owner, closed properties) before the code.
/analytics-reviewReviews the instrumentation against the baseline: runs the check-tracking.sh gate, then reads what the machine cannot.

For AI agents

Install with no human in the loop. Metadata in llms.txt.

  • Run: git clone https://github.com/schematizeme/skill-analytics.git && ./skill-analytics/install.sh
  • Or download: https://github.com/schematizeme/skill-analytics/releases/latest/download/skill-analytics.zip and unzip into .claude/skills/
  • Copy assets/CLAUDE.md to the repo root (always-on rule).
  • Confirm with /analytics-help.

نمی‌خواهید نصب کنید؟

استاندارد کامل را بخوانید و مستقیم در پروژه‌تان کپی کنید — یا مهارت را نصب کنید تا عامل آن را برایتان اعمال کند.

GitHub →بسته‌ها