agentleFS
Sign inStart free
gomarble-aigomarble-ai/marketing-agent/skills/clean-wasted-spend/SKILL.md

clean-wasted-spend

Skill15 starsChanged 37 days agoMITView on GitHub

Use when the user wants to find or cut wasted ad spend: bad search terms, negative keywords, money-losing placements, audiences, devices or locations, fatigued ads still spending, duplicated targeting, zombie campaigns, or Shopping products that never convert. Quantifies each leak, separates real waste from deliberate tests, and proposes the exclusions, negatives and pauses for approval.

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:

    Ask about exceptions: tests, new launches still learning, brand terms and campaigns that are there for reach.

    line 21
  2. Very specific:

    Before recommending any pause, read meta-custom-event-interpretation for Meta sales campaigns that optimize for COMPLETE_REGISTRATION.

    line 37
  3. Very specific:

    Every change: dry run → show each item with its spend and reason → the user approves in chat (they can approve some and reject others) → …

    line 61

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Clean wasted spend

Find wasted spend before the next budget review. Find spend with no path back to target, show the evidence, and approve the clean-up.

When to use

  • "Where are we wasting money?"
  • "Find bad search terms and add negatives."
  • "Which placements, audiences or locations aren't converting?"
  • "Which ads should we pause?"
  • "Clean up the account before the monthly review."

Before you start

  • Get the target (CPA or ROAS) and the date range. Default to 30 days; search terms need the full window.
  • Ask about exceptions: tests, new launches still learning, brand terms and campaigns that are there for reach. recall_memory may know them.
  • Waste means spend that has no credible path back to target. A test inside its learning window isn't waste. Something with too few conversions to judge isn't waste yet; say it's too early.

Where to look

LeakHow to find itMethodology
Search terms (Google)google_ads_run_gaql on search_term_view over the full window, sorted by cost, not just the top performers. Classify each term Q1–Q5.google-ads-search-analysis
Search terms and keywords (Microsoft)bing_ads_get_performance at keyword level, and bing_ads_list_keywords for bids and quality score.Same Q1–Q5 logic
Shopping productsProduct-level performance. Classify KILL / DOWNGRADE / PROMOTE.google-ads-shopping
Placements, devices, locationsBreakdowns on Meta insights (facebook_get_adaccount_insights with breakdowns) and Google segments. Only judge segments with at least 50 clicks and 5 conversions.meta-depth-of-analysis, google-ads-depth-of-analysis
AudiencesAd sets or ad groups spending with poor results; overlapping audiences competing with each other.meta-performance-analysis
Fatigued creativeAds still spending while hook rate, CTR and conversion rate decay.analyse-creative
Zombie campaignsCampaigns with sustained spend and no conversions, well past their learning window.Guardrails skill for the platform
TikToktiktok_get_basic_report_enhanced at ad group and ad level, and tiktok_get_audience_report for segments.tiktok-ads

Before recommending any pause, read meta-custom-event-interpretation for Meta sales campaigns that optimize for COMPLETE_REGISTRATION. Also never pause the top-converting ad in an ad set (meta-guardrails).

Output

  1. Total waste: "$6.1K in the last 30 days has no credible path back to target."
  2. Leaks: a table with the exact query, placement, ad, audience or campaign, its spend and results against target, and the proposed fix.
  3. Held back: items that look bad but are tests, still learning, or too small to judge.
  4. Where the savings go: optionally, which proven campaigns could absorb the freed budget (shift-budget).

Making the change

FixTool
Negative keywords, shared listgoogle_ads_propose_create_negative_keyword_list (new list, optionally attached to campaigns) or google_ads_propose_update_negative_keyword_list (add or remove keywords, attach or detach campaigns)
Negative keywords, one campaigngoogle_ads_propose_update_campaigns
Pause keywords or change match typesgoogle_ads_propose_update_adgroups (needs criterion IDs from GAQL)
Exclude a device, or cut a location's bidgoogle_ads_propose_update_bid_modifiers (a device modifier of 0 excludes it)
Pause Google ads, ad groups or campaignsgoogle_ads_propose_update_ads, google_ads_propose_update_adgroups, google_ads_propose_update_campaigns
Pause Meta adsfacebook_propose_update_ads
Exclude placements or audiences, or pause Meta ad setsfacebook_propose_update_adsets. Targeting is a full replacement, so send the complete targeting with the change.
Pause Meta campaignsfacebook_propose_update_campaigns
TikTok (beta)tiktok_propose_update_ads, tiktok_propose_update_adgroups, tiktok_propose_update_campaigns
Microsoft AdsRead-only through the connector. Give the user the exact negatives and pauses.

Every change: dry run → show each item with its spend and reason → the user approves in chat (they can approve some and reject others) → mode: "live" with only the approved operation_ids. Never go live without that yes.

Put it on a schedule

Offer a weekly search term analyzer or budget leak agent. With write access and ask_approval, it proposes the negatives and pauses each week for approval (see automate-with-agents).

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.

Check your setup, free