Skip to main content

Documentation Index

Fetch the complete documentation index at: https://adadvisor.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

The AdAdvisor audit skill is what a senior media buyer runs on day one of taking over a new Meta (Facebook) ad account. Not “what’s the ROAS?”, but a structured audit: measurement, structure, performance, red flags, in that order. Every performance read is meaningless until you trust the data, and every recommendation is biased until you understand the structure that produced the results.

​
When should I use the audit skill?

Say any of these to your AI and the skill loads automatically:
  • “Audit my account”
  • “Take over this account”
  • “What’s wrong with my account”
  • “Review my ads”
  • “I just inherited an account”
  • “Scan my account”
  • “Give me an overview”
Best for full-account reviews. For investigating one specific campaign, use diagnose instead.

​
How does the AdAdvisor audit work?

1

Signal integrity

Runs adadvisor:get_pixel_health and scores six dimensions: pixel count, last-fired time, server-side event ratio (CAPI coverage), automatic matching fields (EMQ proxy), hosts match with the storefront URL, and AEM priority ordering. Anything red here pauses the rest of the audit. Broken signal makes downstream numbers unreliable.
2

Account structure

Pulls every campaign with adadvisor:list_campaigns and every ad set with adadvisor:list_adsets, then scores: number of active campaigns, ad sets per campaign, undersized ad sets (below the target_CPA × 50 / 7 learning-phase threshold), CBO vs ABO mix, ASC presence with proper exclusions, and naming-convention consistency.
3

Performance baseline

Calls adadvisor:get_performance at multiple windows (7-day for diagnosis, 14-day for decisions, 28-day for trends) and adadvisor:get_timeseries on active ad sets. Each campaign is bucketed into Profitable (ROAS ≥ 1.3x break-even), Marginal (0.8x to 1.3x), or Bleeding (below 0.8x) using your business break-even from the foundation skill.
4

Red flag scan

Cross-references the gathered data against a 15+ pattern detection list: winning ad set capped at low budget, high frequency on small audience, all spend going to one ad, missing exclusion of purchasers from prospecting, wrong optimization event, manual placements restricted, ASC and manual cannibalization, EU ad sets missing DSA fields, and more.
5

Structured report

Produces a five-section report (TL;DR with top 3 fixes, Signal Integrity, Structure, Performance, Red Flags) and ends with “Want me to act on any of these?”. Diagnosis first, mutations on confirmation.

​
What does an AdAdvisor audit report look like?

The audit follows a fixed template so it is predictable across accounts and over time: 
# Account audit: {Business Name}

**Account:** `881523100949350` (USD, America/New_York)
**Date range analyzed:** 2026-04-13 to 2026-05-11 (28 days)
**Total spend (period):** $1,820.04

## TL;DR
- Status: 🟡 YELLOW
- Critical issues: 2
- Top fixes (priority order):
  1. Bump active CBOs from $10 a day to $30 a day. Currently below the exit-learning threshold.
  2. Add CAPI server-side firing on Subscribe and CompleteRegistration events.
  3. Sweep 24 paused legacy TEST campaigns.

## Signal integrity
- Pixel: 🟢 GREEN
- BROWSER:SERVER ratio: 2.8% server, 🟡 YELLOW
- ...

[... structure, performance, red flags sections ...]

## Next steps
- Increase RET CBO from $10 to $30 a day
- Sweep paused TEST campaigns
- Deep-dive RET CBO trajectory

What would you like to act on?

​
What anti-patterns does the audit prevent?

Performance reads will be empty or stale. The skill tells you to wait and pauses the audit instead of producing wrong numbers.
Frequency 3 on a 50 million person audience is fine. Frequency 3 on a 100K person audience is fatigue. The skill always pairs them.
Advantage+ Sales Campaigns have different optimization, learning, and reporting. The skill notes ASC separately and flags cannibalization risk when ASC and manual run on the same audience.
The audit’s job is diagnosis. Mutations come on user confirmation. The skill uses “I recommend X” language, never “I will do X”.
A broken pixel with modeled conversions can look fine for weeks before crashing. The skill surfaces pixel issues as the first finding, not buried under campaign metrics.

​
How do I install the audit skill?

Claude Code

Ships with the full plugin install.

Claude.ai & Desktop

Download adadvisor-audit.zip and upload via Settings, Capabilities, Skills.

​
Frequently asked questions

Install the adadvisor-audit skill (and the foundation skill, which it depends on). Then say “audit my account” or “review my Meta ads” to Claude. The skill calls adadvisor:get_pixel_health, adadvisor:list_campaigns, adadvisor:get_performance, and adadvisor:get_timeseries in sequence, scores each dimension against senior-buyer benchmarks, and returns a structured TL;DR plus a priority-ranked fix list.
Four phases. First, signal integrity (pixel, CAPI, EMQ, AEM priority). Second, account structure (campaign and ad-set counts, CBO vs ABO mix, ASC handling, naming conventions). Third, performance baseline at multiple windows (7-day, 14-day, 28-day). Fourth, red-flag pattern detection (undersized winners, frequency creep, missing exclusions, learning-limited ad sets).
A senior-buyer audit on a new account is one-shot. After that, a daily 15-minute spot-check covers pacing, anomalies, and frequency. A full audit is worth running monthly or whenever you onboard a new team member to the account, run into unexplained performance drift, or take over a client account.
The audit skill is account-wide. It surveys every campaign, scores each one, and produces a prioritized fix list. The diagnose skill is single-entity. It applies the hook-rate to hold-rate to CTR to CVR diagnostic tree to one specific campaign or ad set you point it at, then recommends kill, hold, or scale.
No. The audit’s job is diagnosis. It produces a recommendation list and ends with “Want me to act on any of these?”. Mutations require your explicit confirmation, then route through the relevant action skill (scale for budget changes, diagnose for kills, creative for refreshes).

Foundation

Auto-loaded first. Sets break-even ROAS and target CPA the audit scores against.

Diagnose

Deep-dive on any single campaign the audit flags as bleeding or marginal.

Scale

Act on any campaign the audit flagged as profitable but undersized.
Last modified on May 13, 2026