What does this skill do, and when should you use it?
This skill is the paid-ads activation-stage conversion-tracking QA tool in Aaron Marketing Skills. It checks event firing, UTM hygiene, declared deduplication rules, attribution-window alignment, and offline/iOS ATT modeled gaps. Using the user's GA4 conversion and traffic-acquisition exports plus one manual test conversion, it produces a pre-flight checklist and reusable UTM/event specification. It builds and fixes measurement plumbing; it does not score ROAS R1/R2 or perform ongoing order-ID reconciliation.
It reads the site or account topic, target platforms, the user's GA4 Conversions export, and Traffic-acquisition source/medium export, then asks the user to complete one real manual conversion and verify its event name, value, and source/medium. It checks event firing, UTM hygiene, cross-platform deduplication, attribution-window alignment, offline conversion import, and iOS ATT modeled gaps, marking each item pass, fail, or needs-input based on evidence. It then builds a UTM naming convention and conversion-event specification, documents platform deduplication and attribution windows, flags modeled gaps, emits a handoff summary, and states whether the tracking plumbing is launch-ready. It writes results to memory only after the user explicitly agrees to save them.
- A paid-media team is about to scale Google and Meta campaigns and needs to verify the conversion data path first.
- A marketer sees different order totals in Meta, Google, and GA4 and needs to isolate UTM, deduplication, or attribution-window issues.
- A growth lead is launching a new campaign and needs a UTM convention and conversion-event specification.
- A team has GA4 exports and can run a manual test conversion but does not have ad-platform or tag-manager API access.
How do you install this skill?
- Do not treat missing GA4 exports or a missing manual test conversion as a pass; retain needs-input or fail.
- This skill performs launch-time signal preflight only. Order-ID reconciliation, duplicate-count quantification, and incrementality analysis require the downstream reconciler.
- Exports and memory reports may contain business or conversion data; confirm redaction, access control, retention, and deletion procedures before saving them.
- UTM, attribution-window, iOS-ATT modeling, and platform deduplication behavior can change; responsible operators should verify current platform settings before launch.
- Local filesystem
Install the repository and use the included skill. On a compatible host, run: npx skills add aaron-he-zhu/aaron-marketing-skills. You can also clone it directly: git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. In Claude Code, run /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, then /plugin install aaron-marketing@aaron. The source does not document a dedicated single-skill installation command for this skill.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/aaron-he-zhu/aaron-marketing-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/ad/activate/conversion-signal-qa" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Before launching or scaling, prompt: QA my conversion tracking before I scale. Platforms: Google + Meta. Here is my GA4 Conversions export and Traffic-acquisition (source/medium) export: [paste/path]. You can also ask: Build me a UTM scheme and event spec for this campaign, then give me a pre-launch tracking checklist I can run myself. Provide the GA4 exports and complete the required manual test conversion. After delivery, the skill asks whether to save the results; it does not write memory without confirmation.
What are this skill's strengths and limitations?
- Covers the core pre-launch measurement checks for event firing, UTMs, deduplication, and attribution windows.
- Runs from the user's own GA4 exports and a manual test, without requiring ad-platform APIs.
- Explicitly flags offline-conversion and iOS ATT modeled gaps instead of silently passing them.
- Keeps signal QA separate from ROAS scoring and ongoing attribution reconciliation.
- It does not decide whether ROAS R1/R2 pass; ad-account-auditor is still required for scoring.
- It does not perform order-ID matching, double-count quantification, or incrementality analysis; those belong to attribution-reconciler.
- It depends on usable exports and a user-performed test conversion; without the test, the item remains needs-input.
- The source does not show testing against a particular live ad account, tag manager, or platform configuration.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Paid Ads Conversion Signal QA this page | 56 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
| Dark Social Attribution Analyst | 59 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| Ad Creative Builder | 60 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| Paid Conversion Value Mapper | 57 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
| Paid Ads Audience Segment Builder | 56 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
Unlike ad-account-auditor, this skill builds and fixes the pre-launch measurement signal, while the auditor scores R1/R2 and the full ROAS quality result. Unlike attribution-reconciler, it only gates the existence and alignment of deduplication rules and attribution windows; attribution-reconciler handles ongoing order-ID reconciliation, double-count analysis, and incrementality.
How did FollowSkills review this skill?
The skill shows least-privilege behavior: it primarily uses user-provided GA4/ecommerce exports and a manual test, does not require ad-platform or tag-manager APIs, treats exports as untrusted, and asks for confirmation before writing memory. Points are deducted because it lacks concrete export redaction, retention, and deletion procedures, runtime memory is not encrypted, rollback is not defined, and external platform measurement rules lack independent source attribution.
The instructions, checklist, and UTM/event templates are internally consistent; missing evidence is explicitly classified as needs-input, with defined fail and needs-input feedback. Points are deducted because this is a static review with no skill-specific key-path tests, reproduced input/output cases, or runtime error handling evidence; complex platform differences are largely deferred to another skill.
Triggers, inputs, outputs, scope boundaries, and handoff are reasonably clear, and the skill can operate from user exports without depending on overseas APIs. Points are deducted because non-fit cases, platform-specific boundaries, and edge conditions are incomplete; Chinese support is partial, with no systematic Chinese examples or evidence of mainland-local platform coverage.
The skill has clear frontmatter, versioning, an Apache-2.0 license, Quick Start prompts, a consistent contract, reference materials, limitation disclosures, and a next-skill path. Points are deducted because the skill does not identify a maintainer, changelog, or explicit update process; the homepage is only a repository pointer, publisher identity is unverified, and shared contract dependencies are referenced rather than locally explained.
The documented output is directly useful: a pre-flight checklist, UTM convention, conversion-event specification, deduplication and attribution-window notes, and a launch-readiness decision. Points are deducted because the files provide templates and procedures but no representative real outputs or third-party execution evidence; useful completion still depends on correct exports and a manually performed test conversion.
The checklist, event-spec rules, repository CI, and tests provide some auditable local evidence and make the intended judgments traceable. Points are deducted because the CI tests cover shared architecture and audit validation rather than this skill's event, UTM, deduplication, or attribution paths; there is no third-party execution evidence or multi-source corroboration.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →