Meta Account Audit

Before you start, ask me about each [input] below — one question at a time, with ready-made answers to pick from; skip whatever you can already tell.

Audit the Meta ad account [account: if not listed, call roasr_list_accounts and let me pick] of [client: if not listed, call roasr_list_clients and let me pick] and score it on the RoASr checklist.

1. **Load the RoASr method.** Call roasr_load_skill('roasr_frame'), roasr_load_skill('meta_analysis') and roasr_load_skill('audit_scoring') and follow them exactly.
2. **Frame the data.** Call roasr_limits, roasr_list_clients and roasr_list_accounts: plan, niche, margin, KPI, account currency, sync status and dataSince. Audit [window: last 30 days (default) | last 14 days | last 60 days | custom]. If the window starts before dataSince, say so and narrow it. If the account sync is failed or never, stop: no conclusions without data. Use accountId on every call below.
3. **Structure.** roasr_metrics_campaigns at campaign level, then at adset level (limit 100). Read the evaluation level and the budget type from the settings and the data, not from names. Under a campaign budget or Advantage+, ad set numbers show where the system put the money, not a ranking: do not tell me to cut an ad set just because its share is small. Flag fragmentation: many ad sets, each below the noise floor from the method.
4. **Frequency and fatigue.** From the adset rows: frequency, CTR and CPM. Fatigue is likely only when high frequency comes together with falling CTR or rising CPM; frequency alone is not a verdict. Never sum reach or frequency across rows or days.
5. **Placements.** roasr_metrics_breakdown by placement, then by platform. Look for a placement with a large spend share and clearly worse CPA than the rest, but only if it clears the noise floor; label it a candidate for exclusion, the decision is mine.
6. **Ad quality.** roasr_signals with kind meta_ad_quality (limit 100): ads with below-average quality, engagement or conversion rate rankings and the spend they carry. Rankings are relative to competing ads, treat them as a hint.
7. **Funnel fork.** roasr_metrics_actions for the window. Purchase buckets mean ecommerce: judge ROAS and CPA. Lead buckets mean lead gen: judge CPA only and say that lead quality is not visible here. Both present: judge each separately. Buckets are not additive.
8. **Cross-check.** roasr_list_findings for this client, Meta accounts only: confirm or contradict each open finding with what you saw.
9. **Score.** Go through the Meta Ads part of the audit_scoring checklist. Mark every check pass, warning, fail or n/a. A check you cannot judge from this data is n/a, never a fail. Give the total score and its band.

**Output:** open with one sentence: score, band and the biggest lever. Then a table by checklist category: category · score · main evidence · n/a count. Then 5–8 actions grouped High / Medium / Nice-to-have, each tied to its evidence and the expected effect.

**Output rules:**
- Recommendations only — never apply changes; RoASr tools are read-only.
- reach and frequency are not additive across days; conversion buckets are not additive.
- Without a margin for the client, talk ROAS and CPA, not profit.
- Under the noise floor, say "too little data" instead of a verdict.
- Answer in the language of this prompt.