Ad Account AuditorSAFE

120 marketing skills as an AI marketing staff — plugin, portable skills, or an 8-bot team across 7 disciplines (narrative, SEO/GEO, social, email, paid, influencer, launch) on one contract, with 8 auditor gates: TALE · CORE-EEAT · CITE · ECHO · SEND · ROAS · STAR · RAMP.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: ad-account-auditor
slug: aaron-ad-account-auditor
displayName: "Ad Account Auditor · 付费广告账户审计"
summary: "付费广告账户审计/ROAS评分"
description: 'Use when auditing a paid ad account for incremental contribution, wasted spend, or measurement integrity before scaling; runs a typed 20-item ROAS profile with verified vetoes and a SHIP/FIX/BLOCK/UNDECIDED gate on own exported data. Not for campaign structure design — use campaign-architect; not for creative production — use ad-creative-builder. 付费广告账户审计/ROAS评分'
version: "20.1.0"
license: Apache-2.0
compatibility: "Claude Code and compatible agent-skill hosts"
homepage: "https://github.com/aaron-he-zhu/aaron-marketing-skills"
when_to_use: "Use when checking whether a paid account or portfolio is safe to launch or scale. Requires normalized own-data outcomes, attribution windows, currency, conversion lag, and business constraints."
argument-hint: "<campaign + outcome exports> <currency/window/lag> [profile]"
allowed-tools: WebFetch
class: auditor
metadata: {"author": "aaron-he-zhu", "version": "20.1.0", "discipline": "ad", "phase": "activate", "geo-relevance": "medium", "hermes": {"tags": ["marketing", "ad", "activate"], "category": "ad"}, "openclaw": {"emoji": "🎯", "homepage": "https://github.com/aaron-he-zhu/aaron-marketing-skills"}}
---

# Ad Account Auditor

Audit one paid-media account or portfolio for incremental contribution and operating quality under declared constraints. Platform-reported ROAS is one input, never the objective or truth set by itself.

## When This Must Trigger

- Before launching, materially increasing spend, or changing a risky bid/targeting strategy.
- When tracking, attribution inflation, unsafe placements, claims, or wasted spend are in doubt.
- When the user requests a ROAS/RQS account audit from their exports.

## Quick Start

```text
Audit this USD account for direct response using 7-day click, 3-day lag, and $120 CAC ceiling.
Run the incremental-profit profile against the holdout and order-ID exports.
```

## Skill Contract

**Reads:** one normalized account/portfolio evidence set. **Writes:** only a permissioned v3 artifact. **Done when:** required context and all 20 states are explicit, vetoes use verified evidence, and scorer output is reported without executing spend changes.

This skill judges. `conversion-signal-qa`, `attribution-reconciler`, `campaign-architect`, `ad-creative-builder`, and `budget-pacing-monitor` build/fix the inputs. Never enable campaigns, change bids, upload audiences, or scale budgets without separate explicit approval.

For a pre-launch account-audit request, use the narrow route `conversion-signal-qa` immediately before this gate. Do not automatically insert `placement-exclusion-manager` or `conversion-value-mapper` between signal QA and the audit; missing placement or value evidence remains Unknown in this run, and those sibling builders become separate remediation only when the user requests them or the completed gate identifies the corresponding finding.

## Data Sources

| Need | Preferred evidence |
|---|---|
| Delivery/spend | Campaign, query, placement, audience, and change-history exports |
| Outcome truth | Deduplicated order/lead IDs from ecommerce, analytics, or CRM |
| Economics | Currency, margin/contribution, CAC/payback constraint |
| Attribution | Platform + own-data timestamps/IDs, normalized windows and lag |
| Safety/claims | Placement report, rendered ad/landing, approved claim/disclosure state from [offer-claims-registry](../../../protocol/offer-claims-registry/SKILL.md) (the paid claims SSOT) |
| Incrementality | Holdout/geo split/causal test, otherwise explicitly labeled proxy |

## Instructions

### Runtime Reads

- `../../../references/auditor-runbook.md`
- `../../../references/scoring-semantics.md`
- `../../../references/roas-benchmark.md`
- `../../../references/runtime-invocation.md`
- `references/auditor-runtime.md`

### Runtime and Setup

Read `../../../references/auditor-runbook.md`, `scoring-semantics.md`, `roas-benchmark.md`, and the ROAS catalog entry. Standalone installs use bundled immutable `references/auditor-runtime.md`; never fetch mutable `main`. Before deterministic calls, follow [`runtime-invocation.md`](../../../references/runtime-invocation.md), resolve `AARON_SKILLS_ROOT="${CLAUDE_PLUGIN_ROOT:-$(git rev-parse --show-toplevel 2>/dev/null || true)}"`, and require the scorer, validator, and typed catalogs. If unavailable, return `score_state: NOT_SCORED` / `score_confidence: not_scored` with no gate verdict or persistent artifact.

Declare profile (`direct-response|prospecting|incremental-profit`), target, currency, attribution window, conversion lag, business constraint, goal, and observation date. If any required context is missing, return `NEEDS_INPUT/UNDECIDED`.

### Evidence and Scoring

1. Normalize currency, windows, IDs, lag, and portfolio scope before comparing metrics.
2. Score all 20 `R1..S5` criteria from the benchmark with source/date/type/confidence.
3. Use Unknown for missing own-data truth, placement exports, or reconciliation. No data is not a veto and cannot be N/A merely because access is inconvenient.
4. Verify vetoes:
   - `ROAS-R1`: instrumentation demonstrably fails the named own-data truth set.
   - `ROAS-R2`: material double-counting/inflation is demonstrated.
   - `ROAS-O1`: material claim/disclosure failure against the `offer-claims-registry` approved state.
   - `ROAS-O2`: applicable platform/restricted-category violation.
   - `ROAS-A1`: placement evidence demonstrates a material safety breach.
5. Run the typed scorer. Report estimated/proxy incrementality as such; do not call platform attribution causal.

## §2 ROAS Worked Examples

- Complete direct-response profile, raw 78, no veto/fail: `DONE/SHIP`, final 78.
- Complete profile, raw 78, one verified R1 failure: `DONE_WITH_CONCERNS/FIX`, final 59.
- Complete profile, verified R1 and R2 failures: `DONE/BLOCK`, raw retained, no final score.
- Missing placement

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.

[![ModelsAtlas audit: Ad Account Auditor](https://modelsatlas.com/skills/aaron-he-zhu/ad-account-auditor/badge.svg)](https://modelsatlas.com/skills/aaron-he-zhu/ad-account-auditor/)

Not your listing, or something out of date? Claim or update your listing.