agentleFS
fourteenwmfourteenwm/ppc-ai-skills/account-diagnostic/SKILL.md

account-diagnostic

Skill72 starsChanged 4 months agoMITView on GitHub

42-point Google Ads account inspection (44 with the local-service preset) — discrete GREEN/YELLOW/RED checks across 14 categories (conversion tracking, pacing, impression share, quality score, search-term waste, keyword health, creative, RSA assets, account settings, PMAX config, negative keywords, placement safety, extensions, video & Demand Gen automation), scored into one overall verdict with estimated monthly waste, then triaged per rules.md. Auto-invoke when user says 'run diagnostic', 'inspect account', '40-point inspection', 'diagnostic for [account]', 'audit [account]', or 'account health check'.

  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Check 44 matters because Demand Gen automation lives on the ad (ad_group_ad.ad_group_ad_asset_automation_settings), not the campaign — …

    line 82
  2. Very specific:

    When asked *why* a check fired, answer from references/check-rubric.md — exact thresholds, not paraphrase.

    line 127
  3. Very specific:

    Read-only. This skill runs GAQL SELECT queries exclusively — it never mutates an account.

    line 183

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

42-Point Account Diagnostic

Like a mechanic's multi-point vehicle inspection, but for Google Ads accounts. Every check lands on an explicit GREEN/YELLOW/RED verdict with a finding, a dollar impact where one can be estimated, and a "do this" action line — no narrative to interpret.

One operator owns the workflow end to end: run the inspection → present the scored report → triage the findings (rules.md) → route each actionable finding to its fix skill. The diagnostic itself never changes an account.

Naming note: 42 checks on a standard run; the local-service preset adds two for 44. "40-point inspection" survives as a trigger because the instrument grew from its original 40 core checks.


Invocation Triggers

Auto-invoke when user says:

  • "Run diagnostic for [account]"
  • "Inspect [account]"
  • "40-point inspection"
  • "Audit [account]" / "account health check"
  • "Account diagnostic"

How to Run

cd account-diagnostic
python scripts/run_diagnostic.py --cid 1234567890

Options

--cid CID              Customer ID (no dashes) — required
--name NAME            Account name (optional, display only)
--days N               Lookback period (default: 30)
--vertical V           Calibration preset: property-management | local-service
                       (default: property-management)
--pacing-threshold N   Pacing tolerance % (overrides the vertical preset)
--sheet-id ID          Optional: write the color-coded Inspection tab into an
                       existing Google Sheet (gspread service-account auth)

Preset choice and the deeper tuning knobs: see rules.md → Preset selection.


The Checks

#CategoryCovers
1-3Conversion TrackingPrimary actions exist, firing recently, no orphans/duplicates
4-6Budget & PacingSpending at all, MTD pacing vs tolerance, projected EOM vs budget
7-9Impression ShareSearch IS, Budget Lost IS ($ opportunity), Rank Lost IS
10-11Quality ScoreAverage QS, low-QS concentration
12-13Search TermsZero-conversion waste %, high-spend zero-conv terms
14-16Keyword HealthServing rate, high-spend zero-conv keywords, match-type mix
17-22Creative & AdsDisapprovals, DKI, auto-created assets, dead URLs, seasonal copy, auto-apply
23-25RSA AssetsPerformance labels, LOW-rated control, BEST-rated existence
26-29Account SettingsLocation targeting mode, content suitability, auto-apply, geo exclusions
30-34PMAX ConfigText/image automation, URL expansion, search themes, audience signals
35-36Negative KeywordsCoverage per campaign, negative-vs-positive conflicts
37-38Placement SafetySuspicious placements, brand-unsafe YouTube placements
39-40ExtensionsCore (sitelinks/callouts), supplemental (snippets/images)
41-42Local Service (local-service preset only)Call extensions, location extensions
43-44Video & DGen AutomationPMAX video enhancements; Demand Gen ad-level asset automation

Per-check GREEN/YELLOW/RED criteria, dollar-impact formulas, and preset thresholds: references/check-rubric.md.

Check 44 matters because Demand Gen automation lives on the ad (ad_group_ad.ad_group_ad_asset_automation_settings), not the campaign — campaign-level audits structurally can't see it. Five settings are inspected per ad (design versions for images, generate-videos-from-assets, vertical conversion, shorter videos, landing-page preview), all but one of which default ON. Pair with the dgen-automation-disable skill to fix what this check flags.


Scoring

  • Overall RED: any auto-red circuit breaker (e.g. no conversion tracking), or 3+ RED checks
  • Overall YELLOW: 1-2 RED, or 6+ YELLOW
  • Overall GREEN: everything else
  • Estimated waste/mo: sum of the dollar impacts across flagged checks (pacing gap and EOM projection overlap — read it as a ceiling, not a precise total)

Checks that don't apply (e.g. PMAX checks on an account with no PMAX) report N/A and don't count against the score.


Vertical Presets

PresetTuned forDifferences
property-managementLead-gen portfolios with strict pacingBaseline: pacing ±8%, seasonal copy & missing geo-exclusions are RED, negative bar 20/campaign
local-servicePhone-driven local businesses (auto repair, home services)Pacing ±10%, seasonal copy & missing geo-exclusions soften to YELLOW, negative bar 10, lower zero-conv dollar floors, adds checks 41-42

After the run — operator duties

The script's exit is the midpoint of this skill, not the end. On every run:

  1. Present the verdict — overall color, the GREEN/YELLOW/RED/N/A counts, and estimated waste (always framed as a ceiling).
  2. List the REDs with their findings and dollar figures, largest first.
  3. Read rules.md before recommending anything — apply the triage order (circuit breakers → dollar REDs → structural REDs → YELLOWs) and rule out the false-alarm patterns. A RED that's a launch artifact gets contextualized, never rewritten.
  4. Route each actionable finding per the rules.md routing table — every fix is its own skill with its own dry-run and human approval, or an exactly-specified UI change. Order fixes by dependency (waste before budget, automation off before creative refresh).
  5. When asked why a check fired, answer from references/check-rubric.md — exact thresholds, not paraphrase. examples.md shows the expected shape of a triage read-out.

What this skill deliberately does NOT do

  • No mutations, ever. GAQL SELECT only. Fixing findings is a separate, human-approved step through mutation-safety-gated skills — see the routing table in rules.md.
  • No portfolio sweeps. One account per run. Choosing which account to inspect first is portfolio-health-prioritization's job.
  • No root-cause investigation. The inspection says what is out of spec; why is the deep-dive skills' job (underspending-investigation, impression-share-diagnostics).
  • No auto-generated budget changes. Pacing findings produce diagnosis plus a conservative recommendation; budget execution belongs to humans.

Files in this skill

FilePurpose
SKILL.mdThis file — how to run + what the operator does with results
README.mdZero-context setup guide: install, prerequisites, usage
rules.mdTriage decision logic — read after every run, before recommending
examples.mdWorked triage reads, incl. a false alarm and a refusal edge case
references/check-rubric.mdPer-check verdict criteria + the 8 preset knobs
scripts/run_diagnostic.pyThe inspection engine (read-only GAQL)

Runtime output lands in data/ (gitignored), never inside the docs.


Output

  1. Console report — grouped by category, with icons, findings, and actions
  2. CSV — full checklist written to data/diagnostic-<cid>-<timestamp>.csv
  3. Google Sheet (optional, --sheet-id) — color-coded Inspection tab written into a sheet you own; requires a gspread service account (share the sheet with the service account's client_email)

Prerequisites

  • google-ads.yaml in the skill folder — see the google-ads-api-setup skill
  • pip install google-ads (plus gspread google-auth if using --sheet-id)
  • Google Ads API access at whatever level your yaml grants (single account or MCC with login_customer_id)

Safety

Read-only. This skill runs GAQL SELECT queries exclusively — it never mutates an account. Fixing what it finds is a separate, human-approved step (see mutation-safety and the routing table in rules.md).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.