---
name: full-audit
description: Complete multi-platform paid media audit orchestrating all specialist agents in parallel. Platform-segmented output with period comparison, change history, campaign-level optimization, and audience verification. ALWAYS invoke this skill when the user says 'full audit', 'complete audit', 'audit everything', 'account health check', 'run a full audit', 'check everything', 'comprehensive review', 'deep audit', or wants ANY broad review of their advertising. Also invoke when auditing a new client for the first time. This project-level skill takes priority over any global ads skill.
model: opus
context: fork
argument-hint: "[client-name] [website-url]"
allowed-tools: Read, Grep, Glob, Bash, Write, Agent, mcp__google-analytics__run_report, mcp__google-analytics__get_account_summaries, mcp__google-analytics__get_property_details, mcp__google-analytics__get_custom_dimensions_and_metrics, mcp__google-analytics__list_google_ads_links, mcp__google-ads__list_accounts, mcp__google-ads__get_campaign_performance, mcp__google-ads__get_ad_performance, mcp__google-ads__get_ad_creatives, mcp__google-ads__run_gaql, mcp__google-ads__execute_gaql_query, mcp__google-ads__get_image_assets, mcp__google-ads__get_asset_usage, mcp__google-ads__get_account_currency, mcp__google-search-console__search_analytics, mcp__google-search-console__enhanced_search_analytics, mcp__meta-ads__get_campaigns, mcp__meta-ads__get_adsets, mcp__meta-ads__get_ads, mcp__meta-ads__get_ad_creatives, mcp__meta-ads__get_insights, mcp__meta-ads__get_custom_audiences, mcp__meta-ads__get_pixels, mcp__meta-ads__get_account_info, mcp__meta-ads__get_ad_accounts, mcp__meta-ads__list_catalogs, mcp__meta-ads__list_product_sets, mcp__claude-in-chrome__navigate, mcp__claude-in-chrome__read_page, mcp__claude-in-chrome__get_page_text, mcp__claude-in-chrome__tabs_context_mcp, mcp__claude-in-chrome__tabs_create_mcp
---

# Full Paid Media & Tracking Audit (V2)

Comprehensive platform-segmented audit of a client's entire advertising performance ecosystem. Orchestrates 11 to 14 specialist agents organized by platform, with mandatory period comparison, change history, campaign-level optimization, and audience verification gates.

## Mandatory Rules

Load these rules BEFORE any analysis. They apply to every agent and every finding:
1. `.claude/rules/period-comparison-protocol.md` (every metric shows 30d vs prev 30d AND 14d vs prev 14d)
2. `.claude/rules/change-history-first.md` (pull change history BEFORE performance analysis)
3. `.claude/rules/recommendation-depth-standard.md` (from/to/rationale/impact/risk for every recommendation)
4. `.claude/rules/exact-entity-names.md` (verbatim API names, never paraphrase)

## Arguments

$ARGUMENTS should include:
- Client name (required)
- Website URL (optional, will be read from client profile)
- Specific focus areas (optional, default is full coverage)
- Known issues or areas of concern (optional)

## Instructions

### Step 1: Client Context Load

1. Check `clients/{client-name}/` for existing profile, history, and open items
2. Read ALL client files: profile.md, history.md, open-items.md, CLAUDE.md, tracking-config.json
3. Pull Asana parent task and ALL subtasks for the client to see current open work
4. Identify all account IDs: Google Ads CID, Meta account ID, GMC Merchant ID, GA4 Property ID, GTM Container IDs
5. Calculate date ranges for period comparison:
   - Current 30D: yesterday minus 30 to yesterday
   - Previous 30D: yesterday minus 60 to yesterday minus 31
   - Current 14D: yesterday minus 14 to yesterday
   - Previous 14D: yesterday minus 28 to yesterday minus 15

### Step 2: Change History and Period Data Collection

BEFORE launching the agent swarm, pull change history from both platforms:

**Google Ads:** Run the change_event GAQL query from `.claude/rules/change-history-first.md` for the last 60 days. Parse into a timeline of what changed and when.

**Meta Ads:** Pull campaign and ad set `created_time` and `updated_time` fields. Note any campaigns or ad sets modified in the last 30 days.

Pass the change history summary and explicit date ranges to every agent in Step 3.

### Step 3: Platform Agent Swarm (Parallel)

Launch all agent clusters simultaneously. Each agent prompt MUST include:
- Client name and all relevant account IDs
- Explicit date ranges for current and previous periods (calculated in Step 1)
- The change history summary from Step 2
- Instruction to use exact entity names
- Instruction to paginate fully for audience pulls

#### Google Ads Cluster (6 to 9 agents)

| Agent | Focus |
|---|---|
| platform-strategist | Campaign structure, settings review, audience assignment per campaign, campaign-level mini audits |
| pmax-specialist | Asset groups, channel splits, audience signals, listing groups, product-level deep analysis, brand rollups |
| keyword-strategist | QS analysis, search terms, negatives, match types, impression share, campaign-level keyword audits |
| tracking-auditor | Conversion actions, enhanced conversions, change history correlation, event parameter verification |
| budget-optimizer | Budget allocation, pacing, IS lost to budget, scaling candidates, bidding strategy evaluation per campaign |
| creative-analyst | RSA performance, ad strength, extensions audit |
| attribution-analyst | Cross-platform attribution, conversion paths, GA4 reconciliation |
| landing-page-analyst | Page speed, mobile experience, conversion flow, message match (if Chrome available) |
| compliance-auditor | Policy violations, disapprovals, restricted categories (if applicable) |

#### Meta Ads Cluster (5 to 8 agents)

| Agent | Focus |
|---|---|
| platform-strategist | Campaign structure, ad set targeting, learning phase, ASC assessment, audience infrastructure (with FULL pagination) |
| meta-creative-analyst | Creative performance, fatigue, format diversity, A+ enhancement audit, placement splits, campaign-level creative health |
| meta-capi-auditor | CAPI health, EMQ, dedup, pixel config, 14 silent mistakes, GTM/Stape integration |
| budget-optimizer | Meta budget distribution, prospecting vs retargeting split, bid strategy alignment |
| attribution-analyst | Meta attribution windows, view through vs click analysis, GA4 cross-validation |

#### Commerce Manager (1 agent, ecommerce clients only)

| Agent | Focus |
|---|---|
| commerce-manager-audit skill | Catalog health via Chrome, product sets, checkout URL, pixel connection, event source groups |

#### Merchant Center (1 agent, ecommerce clients only)

| Agent | Focus |
|---|---|
| feed-diagnostician | Feed health, disapprovals, custom labels, product-level deep analysis, margin tier performance, zombie products |

#### Tracking Deep Dive (1 agent)

| Agent | Focus |
|---|---|
| gtm-tracking-deep-dive skill | GTM API tag audit, Stape container health, Chrome live verification, click ID persistence |

### Step 4: Audience Verification Gate

After ALL agents return, run a dedicated verification step:

1. Re-pull ALL Google Ads audiences via `user_list` GAQL query without LIMIT
2. Re-pull ALL Meta custom audiences via `mcp__meta-ads__get_custom_audiences` with explicit pagination (follow `after` cursor until no more pages)
3. Cross-reference against agent reports:
   - If an agent claimed an audience type "does not exist" but the verification gate finds it, flag the discrepancy and correct the finding
   - If the verification gate finds audiences not mentioned by any agent, flag them as "discovered but unanalyzed"
4. For each audience found: record exact name, type, size, and which campaigns use it

### Step 5: Report Synthesis

Map all agent findings to the platform-segmented template in `templates/deep-audit-report.md`:

**Section 1: Google Ads Audit & Strategy**
- Account overview with period comparison table
- Change history summary
- Campaign-by-campaign analysis (every enabled campaign gets its own subsection)
- PMax deep dive
- Search deep dive
- Shopping/feed performance
- Prioritized recommendations (from/to/rationale/impact/risk format)

**Section 2: Meta Ads Audit & Strategy**
- Account overview with period comparison table
- Change history summary
- Campaign-by-campaign analysis
- Creative health
- Audience infrastructure (COMPLETE inventory with exact names and assignment verification)
- CAPI & signal quality
- Prioritized recommendations

**Section 3: Meta Commerce Manager** (ecommerce only)
- Catalog health, product sets, checkout URL, pixel connection

**Section 4: Google Merchant Center** (ecommerce only)
- Feed health, custom labels, disapprovals, product-level analysis

**Section 5: GTM & Tracking Infrastructure**
- GTM container audit, Stape health, pixel firing verification, click ID persistence

**Section 6: Cross-Platform Analysis**
- Attribution reconciliation (Google vs Meta vs GA4 vs backend)
- Budget portfolio analysis with marginal efficiency
- Competitive landscape observations
- Seasonality patterns
- Creative testing framework recommendation
- Audience expansion opportunities
- Automation opportunities

**Implementation Roadmap**
- Phase 1: Critical (Week 1 to 2)
- Phase 2: High Impact (Week 3 to 4)
- Phase 3: Optimization (Month 2)
- Phase 4: Advanced & Ongoing (Month 3+)

### Step 6: Proactive Insights

After all analysis is complete, generate additional insights the client did not ask for:
- Competitive landscape observations from auction insights and search term data
- Seasonality patterns from historical performance (are there predictable peaks/troughs?)
- Creative testing framework for the next quarter (what to test, in what order, with what budget)
- Audience expansion opportunities (untapped segments, cross-platform audience strategy)
- Automation opportunities (scripts, rules, automated reports)
- Upcoming platform changes that affect this account (API deprecations, feature rollouts, policy changes)

### Step 7: Save Results

1. Save the full report to `clients/{client-name}/reports/{date}-deep-audit.md`
2. Generate Word document via docx skill
3. Update `clients/{client-name}/history.md` with session summary
4. Update `clients/{client-name}/open-items.md` with new action items
5. Create Asana subtask under client parent with audit summary
6. Add verification comment to the Asana subtask with: what was audited, how many agents deployed, key findings count, data sources used

## Output Verification

Before presenting the audit to Michael, confirm:
- All agents returned findings (no silent failures). If any agent failed, note which queries were unexecuted.
- Every metric table has period comparison columns (30d vs prev 30d, 14d vs prev 14d)
- Change history section appears before performance findings in every platform section
- Every recommendation follows the from/to/rationale/impact/risk format
- Every entity name is verbatim from the API (no paraphrased campaign or audience names)
- Audience verification gate ran and no false absence claims remain
- Report is organized by platform section (not by function)
- Commerce Manager section is populated for ecommerce clients
- GTM/tracking section includes programmatic tag audit results
- Implementation roadmap has specific dollar impact projections
- Client memory (history.md, open-items.md) was updated
- Report saved to the correct client directory

## Output

The final deliverable is a platform-segmented audit report suitable for direct handoff to the OFM team, with:
- Each platform section self-contained (team members find their section without reading everything)
- Campaign-level optimization recommendations with quantified rationale
- Period comparison showing trend direction on every metric
- Change history explaining recent performance shifts
- Proactive insights beyond what was asked
- Implementation roadmap with specific dollar impact and effort estimates per action item