# Tracking Verify

> Use when the user wants preview QA, failure interpretation, release readiness, or an explicit publish handoff.

- Skill: `jtrackingai/tracking-verify` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add jtrackingai/tracking-verify`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jtrackingai/tracking-verify/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: jtrackingai (https://skillmd.com/u/jtrackingai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jtrackingai/tracking-verify

---


# Tracking Verify

Use this skill for verification and optional publish handoff.

## Inputs

- `<artifact-dir>/event-schema.json`
- `<artifact-dir>/gtm-context.json`

## Workflow

If the telemetry consent prompt appears and no prior choice is recorded, stop and follow [../../references/telemetry-consent.md](../../references/telemetry-consent.md) before continuing.

Run preview:

```bash
./event-tracking preview <artifact-dir>/event-schema.json --context-file <artifact-dir>/gtm-context.json
```

If comparing against an older preview run, pass the previous health baseline:

```bash
./event-tracking preview <artifact-dir>/event-schema.json --context-file <artifact-dir>/gtm-context.json --baseline <previous-tracking-health.json>
```

`preview` launches a real Chromium via Playwright and exercises the live site to fire GA4/GTM events for verification. Run it in an environment that permits outbound network and local browser execution; environments that restrict either tend to cause Playwright to hang or fail silently rather than return a clean error.

Then interpret:

- blockers
- expected failures
- selector mismatches
- unexpected fired events outside the approved schema
- release readiness

If the user explicitly wants to publish after verification:

```bash
./event-tracking publish --context-file <artifact-dir>/gtm-context.json --version-name "GA4 Events v1 - <date>"
```

If `tracking-health.json` is missing, still manual-only, or has blockers, `publish` now stops by default. Only use `--force` when the user explicitly wants to override that gate.

## Required Output

Produce and share:

- `<artifact-dir>/preview-report.md`
- `<artifact-dir>/preview-result.json`
- `<artifact-dir>/tracking-health.json`
- `<artifact-dir>/tracking-health-report.md`
- `<artifact-dir>/tracking-health-history/`
- updated `<artifact-dir>/workflow-state.json`

## Closeout Style

- default to a verification verdict first: healthy, blocked, or manual follow-up required
- summarize blockers, unexpected events, and release-readiness in plain language before listing files
- keep raw preview data and artifact references after the summary

## Stop Boundary

- stop after preview if the user only asked for QA
- publish only when the user explicitly wants to affect the live site

If the platform is Shopify, switch to the Shopify-specific rules in `tracking-shopify`.

## References

- [../../references/preview-report.md](../../references/preview-report.md)
- [../../references/gtm-troubleshooting.md](../../references/gtm-troubleshooting.md)
- [../../references/output-contract.md](../../references/output-contract.md)