agentleFS
EricStrohmaierEricStrohmaier/adchat-mcp/skills/diagnose-performance/SKILL.md

diagnose-performance

Skill1 starsChanged 3 days agoMITView on GitHub

Find out why Meta or Google Ads results changed. Use when the user says results dropped, CPA or cost per result went up, ROAS fell, spend stopped, "what happened to my ads", "why is it worse this week", or asks for a health check of a Meta (Facebook/Instagram) or Google Ads account through AdChat.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. gomarble-ai/marketing-agent★ 15same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository8 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Look at spend, results, cost per result, CTR, CPM (Meta) or CPC (Google).

    line 24
  2. Very specific:

    Read-only until the user picks a fix.

    line 59
  3. Distinctive:

    Don't call a change a trend on fewer than 3 days of data or a handful of conversions.

    line 60
  4. Distinctive:

    Don't recommend raising budget to fix a drop you haven't explained.

    line 64

≈ “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

Diagnose performance

Goal: find the one or two real reasons results changed, backed by numbers, and suggest the smallest fix. Most "my ads stopped working" cases are one of: delivery stopped, tracking broke, the creative tired out, the audience got too small or too expensive, or the comparison is unfair (a holiday, a sale, too few days of data).

Steps

  1. Context first. get_account_briefing for the account's goals, target CPA or ROAS and notes. If the user has no target, ask what a good result costs them.
  2. Pick the account. meta_list_accounts (or google_get_accounts). Work in one account at a time.
  3. Confirm the change is real. Compare two equal windows, for example the last 7 days against the 7 before.
    • Meta: meta_insights at level: account with time_increment: 1 (daily) over 14 to 28 days.
    • Google: google_get_insights with breakdown: date. Look at spend, results, cost per result, CTR, CPM (Meta) or CPC (Google). Small numbers (under ~20 conversions a week) swing a lot: say so.
  4. Find where it moved. Drill down one level at a time, sorted by spend:
    • Meta: meta_insights at level: campaign, then adset, then ad.
    • Google: google_list with level: campaigns, then ad_groups, keywords. Usually one campaign or ad set explains most of the change.
  5. Check the usual causes, in this order:
    • Delivery: meta_diagnose with kind: delivery_issues (disapproved ads, learning limited, budget or billing problems). Google: google_list with level: ads for disapprovals and status.
    • Tracking: meta_diagnose with kind: pixel_health. A sudden drop to near zero conversions with steady clicks is usually tracking, not ads.
    • Creative fatigue: frequency above ~3 to 4 in prospecting with CTR falling week over week. See the creative-analysis skill.
    • Cost pressure: CPM up with CTR flat means the auction got pricier (seasonality, competition), not that the ads got worse.
    • Breakdowns: meta_insights with breakdowns such as age, publisher_platform, platform_position or country to spot one segment eating budget. Google: google_get_insights with breakdown: device or network.
    • Recent edits: get_change_history (Meta) or google_get_change_history. A budget jump or targeting edit often resets learning.
  6. Check money, if connected. get_true_profit compares Meta spend with real revenue; it can show that a "worse" ROAS is actually fine, or the reverse.
  7. Open alerts. get_alerts may already describe the problem and a suggested fix.

What to show the user

  • One sentence answer first: "Cost per purchase rose from $18 to $27, mostly because ad set Broad US lost its best ad to a disapproval on Tuesday."
  • A small table: the before/after window for the 3 to 5 numbers that matter.
  • The cause, with the evidence, and how sure you are.
  • One to three fixes, smallest first. Name what each will change.

Guardrails

  • Read-only until the user picks a fix. Fixes go through the safe-changes flow.
  • Don't call a change a trend on fewer than 3 days of data or a handful of conversions.
  • Meta and Google count conversions differently; don't add them up as if equal.
  • Meta budgets are in cents; insights spend is in currency units. Google amounts are in currency units.
  • Don't recommend raising budget to fix a drop you haven't explained.

Example prompt

"My Facebook cost per lead doubled this week. What happened?"

Briefing → account-level daily insights for 14 days → campaign and ad set level → meta_diagnose delivery issues → frequency and CTR by ad → change history. Answer: "One ad set's frequency hit 5.2 and CTR halved; the others are steady. Pause its two oldest ads and add fresh creative. Want me to pause them?"

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.