agentleFS
gomarble-aigomarble-ai/marketing-agent/skills/google-ads-guardrails/SKILL.md

google-ads-guardrails

Skill15 starsChanged 38 days agoMITView on GitHub

Guardrails for Google Ads recommendations and mutations. Hard rules on what NOT to recommend, prohibited phrasing, data-quality requirements. Used internally for any Google Ads optimization or change recommendation.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    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:

    If a recommendation does not map to one of these controls, it is not actionable.

    line 25
  2. Distinctive:

    Ecommerce: Use ROAS, conversion value, cost per purchase.

    line 40
  3. Distinctive:

    Lead gen: Use CPA, cost per lead. Never use ROAS unless offline revenue is imported.

    line 41
  4. Distinctive:

    Unclear: ASK the user. Do not assume.

    line 42
  5. Distinctive:

    Never scale if: CPA > target, ROAS < target, or query/search term hygiene not done

    line 47
  6. Distinctive:

    Never scale based on: Impression Share alone, CPC alone, or early data (< 7 days)

    line 48
  7. Distinctive:

    Do not stack changes (budget + bids + structure simultaneously)

    line 53
  8. Distinctive:

    Do not judge performance on: < 30 conversions (Search), < 50 conversions (PMax/Shopping)

    line 54
  9. Distinctive:

    Avoid changes that reset learning unless absolutely required

    line 55
  10. Very specific:

    Before any scale/pause/restructure suggestion, check:

    line 69
  11. Distinctive:

    Search term quality (Q4/Q5 presence)

    line 70
  12. Distinctive:

    If diagnostics are incomplete → say insufficient data, do not recommend.

    line 77

≈ “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

Google Ads Guardrails

READ BEFORE GENERATING ANY RECOMMENDATION.

Budget Reallocation Prohibition

You CANNOT allocate, shift, increase, reduce, or "optimize" budget for: keywords, search terms, match types, devices (unless bid adjustments in Search), locations (unless bid adjustments or exclusions), audiences (unless bid modifiers), asset groups (PMax has no asset-group budgets).

Before writing ANY recommendation, ask: Does this imply moving money between keywords, queries, products, devices, audiences, or asset groups? If YES → rewrite using valid controls below.

Valid Budget Controls

Campaign TypeValid Controls
Search / ShoppingCampaign daily budget, bid strategy targets (tCPA/tROAS/max CPC), pause/enable campaigns or ad groups
Performance MaxCampaign daily budget, bid strategy targets (tCPA/tROAS), pause asset groups or campaign

If a recommendation does not map to one of these controls, it is not actionable.

Mandatory Rewrites

❌ "Shift budget from keyword A to keyword B" ✅ "Pause keyword A. Increase campaign daily budget by 10–20% to allow keyword B more eligible demand."

❌ "Allocate more to [breakdown: placement/device/demographic/product/asset group]" ✅ "Create a new campaign targeting [dimension] with its own budget, OR use bid adjustments/exclusions."

Rule: Any recommendation implying budget movement between sub-campaign entities must be rewritten to use campaign budget, bid targets, pause/enable, or structural separation.

Metric Selection

First infer conversion type:

  • Ecommerce: Use ROAS, conversion value, cost per purchase. Never use CPA alone for revenue decisions.
  • Lead gen: Use CPA, cost per lead. Never use ROAS unless offline revenue is imported.
  • Unclear: ASK the user. Do not assume.

Scaling Limits

  • Max budget increase: 20% at a time
  • Never scale if: CPA > target, ROAS < target, or query/search term hygiene not done
  • Never scale based on: Impression Share alone, CPC alone, or early data (< 7 days)
  • Frame scaling as: "Incremental test with downside risk"

Learning & Stability

  • Do not stack changes (budget + bids + structure simultaneously)
  • Do not judge performance on: < 30 conversions (Search), < 50 conversions (PMax/Shopping)
  • Avoid changes that reset learning unless absolutely required

Immediate STOP Triggers

Halt any recommendation that attempts:

  • "Scale keywords / products / asset groups"
  • "Reallocate spend between queries"
  • "Optimize placements in PMax"
  • "Increase bids and budget at the same time"
  • "Scale > 20% in one step"
  • "Fix low IS by increasing budget blindly"

Required Diagnostics Before Any Action

Before any scale/pause/restructure suggestion, check:

  1. Search term quality (Q4/Q5 presence)
  2. CPA/ROAS vs target
  3. IS Lost (Rank vs Budget breakdown)
  4. CPC trend direction
  5. Conversion volume sufficiency
  6. Bidding strategy compatibility

If diagnostics are incomplete → say insufficient data, do not recommend.

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.