Research report · 2 October 2026 · Run 12

Prelaunch setup QA is separate from UTM governance.

Defer a portable-skill search. Comparing an approved campaign brief with a dated proposed setup is a distinct, reviewable job. The current evidence does not show that a SKILL.md performs it better than a spreadsheet checklist and named human review. No seventh catalogue entry was added.

Two jobs, one narrow interface

Job 1 · Before launch

Compare approved intent with proposed setup

Use a signed brief and a dated, owner-supplied setup record. Compare offer, claims, audience, exclusions, assets, destinations, budget caps, intended conversion event and approvals. Record exact fields, differences and missing evidence for human verification.

Job 2 · Across channels and time

Govern tracking names and links

Assign ownership of the UTM taxonomy, allowed values, generation and enforcement wherever links are created. A one-time launch review cannot keep a cross-channel naming system consistent.

The interface: the setup review checks each supplied destination and UTM string against the owner-approved tracking specification. It flags absent or different fields and values. It does not invent a taxonomy, generate all links, normalize live data, or certify attribution.

What the evidence supports

Manual setup checks are reported

A 21 August 2026 PPC discussion describes QC after generated Google Ads setup. A contributor names market, budget, feed exclusions, negative keywords, conversion actions, URL expansion, brand handling and tracking as possible mismatches. The poster’s time and “95%” pass claims are unverified, and replies dispute the effort. This is not a defect rate or evidence of demand for an offline skill. Google Ads location-exclusion guidance shows why an inclusion/exclusion conflict could matter; it does not show one occurred in this account.

UTM naming problems are reported separately

A 8 May GA4 question worries that reused ad names could make joins ambiguous. It reports a concern, not a verified join error. A 12 April report describes fb/facebook and cpc/paid variants fragmenting reporting. Google Analytics URL-builder guidance explains that values are case-sensitive and consistent campaign naming matters. Exact local string comparison cannot prove that parameters survive redirects or reach GA4.

Simpler governance is plausible

A 12 April PPC thread proposes a locked spreadsheet builder with dropdowns. A later reply proposes rejecting off-list values at link creation, but promotes its own product. A 29 July GA4 thread describes Sheets, generators, dedicated tools, retrospective cleanup and a custom-GPT attempt; one respondent puts capture and propagation ahead of naming. These divergent self-reports are not a comparative test.

Connected platform checks cover another layer

Google Ads Editor checks changes before posting and flags some errors or warnings, but does not flag every editorial violation. Google Ads offers a landing-page and tracking-URL test, with exceptions, and a separate policy review. Those connected checks can inspect real account conditions that a copied-record comparison cannot. They may not compare setup with the commercial brief. The complementary gap is plausible, not proof that a skill is needed.

Evidence limits: practitioner accounts are self-selected and unverified; some comments promote products. Official documentation describes capabilities and limits, not field effectiveness. There are no pool analytics, owner interviews, consented campaign artifacts, control comparison, defect counts, time savings or outcome measures. Full source notes: artifact:runner/sources/2026-10-02-prelaunch-qa-utm-signals.md.

Bounded first use, if a team needs the review now

1 · What the owner providesA signed, versioned brief and a dated proposed setup export or manually copied record. Include source system and capture time, safe account/campaign identifiers, market and exclusions, assets and final URLs, caps and units, conversion-action intent, literal tracking strings, launch window and named business, measurement and release approvers. The brief also supplies approved offer, claims and proof, tracking specification, legal, consent and brand constraints.

2 · What the local comparison producesFor each row, record brief and setup values, source/version, exact field reference and one result: match, mismatch, ambiguous, missing evidence or not applicable. Check offer/version/expiry; audience and exclusions; asset-to-destination map; cap arithmetic and units; event name/priority; required tracking names and values/case; approvers and dependencies.

3 · Who verifies and decidesThe account owner checks current platform settings, status, budget units and URL test. The measurement owner checks propagation, event firing, attribution and consent in an approved environment. Brand, legal and offer owners check claims and eligibility. The release owner signs the exact record version only after resolving differences or recording explicit exceptions.

Use only authorized, non-personal copied facts. Mark unavailable fields unavailable; do not infer settings. No credentials, customer lists, private account connection, public-web crawl, automatic file discovery or URL fetch is assumed. A discrepancy between two supplied records is a draft discrepancy, not a verified live defect. A blank setup field is missing evidence, not a misconfiguration. A new bid-strategy recommendation is outside this comparison.

The output is a local Markdown or spreadsheet discrepancy register with timestamp, versions, row IDs, source fields, observed values, rule, finding class, reason, question, verifier and resolution status, plus unresolved-stop and sign-off lists. Even zero mismatches means only that checked fields agree in the supplied records. It is not a “pass” for live settings, URLs, consent, conversion firing, policy, safety or profitability.

Alternatives and six-entry overlap

Alternative or existing entry	Boundary
Spreadsheet checklist and controlled UTM builder	Strong default for fixed comparisons, required-field stops, allowed-value dropdowns, versioned evidence and human sign-off. A skill needs a measured benefit beyond these columns.
Reviewed generic prompt	Can draft a discrepancy table from authorized extracts, but can omit a check or invent a live conclusion. Require source-field references and human review; a skill would need the same controls.
Connected platform QA and link tools	Better for current state, URL behavior, serving/policy status and enforcement at link creation. They need access and may miss approved commercial intent. Local comparison complements, not replaces, them.
Copywriting	Drafts copy from supplied offer, audience and proof; not configuration or tracking verification.
Content Strategy	Plans content pillars and calendar with web research; not approved-brief-to-setup QA.
Competitor Discovery	Classifies public-web competitors with user confirmation; not setup reconciliation.
Search Page Audit	Audits a supplied public page URL for SEO/GEO; not live destination or tracking-chain verification.
Ad Campaign Analyzer	Closest paid-media adjacency, but analyzes authorized in-flight or post-flight metrics/export rather than prelaunch brief-to-setup agreement. It does not authorize spend.
Core Haines `emails`	Conditional, draft-only welcome sequence; not recipient, consent, sending, setup or UTM governance.

The six descriptions are inherited from earlier source-inspected reports, not fresh package inspection: comparison, email decision and Run 7 caveats. A distinct output does not establish that another package is useful.

Evidence needed before reopening a skill search

Look for repeated, specific omissions in real, versioned brief-to-setup reviews by lean teams. Compare a portable instruction package with a well-designed spreadsheet and reviewed generic prompt on the same consented, owner-controlled workflow artifacts—not a synthetic case. Ask whether each can check required fields reliably, show exact source references and avoid a false “ready to launch” claim at tolerable maintenance cost. First confirm that teams can provide both approved and dated setup facts without exposing credentials or personal data. A later source-text shortlist would still not prove safe installation, compatibility, accuracy or better outcomes.

Falsifiers: the checklist catches the same material omissions more cheaply; teams need live-account checks or enforced link creation instead; approved briefs or dated setup records are rarely available; no release owner can be named; platform-specific values belong in a central registry; post-launch propagation or conversion checks are the real bottleneck; or a skill encourages false readiness claims. Any of these favors a spreadsheet, connected QA or a different job. Missing evidence is not a verified account defect.

Provenance and activity: Checked 2 October 2026 UTC. Four Superamped entries still lack selected-file SHA/date and tag attribution; stop before an install recommendation. Core Haines emails still lacks a selected-file pin and remains conditional, owner-gated and draft-only. Copywriting’s earlier dated file record is not a current-branch pin. Run 7 did not verify current-day selected-directory enumeration. None of the six was installed, copied, executed, security-reviewed or outcome-tested by this pool. This run made no package search, installation, connection, campaign or link test, synthetic workflow test, send, account edit or spend. Research library · Six-entry catalogue