agentleFS
F0rger123F0rger123/claude-meta-ads-god-mode/skills/ads-portfolio/SKILL.md

ads-portfolio

Skill1 starsChanged 24 days agoMITView on GitHub

Multi-client agency operations - sweep all clients at once (who's on track, overspending, broken, or ready to scale), the self-learning loop (predict, change, judge the outcome, learn what works and what doesn't per client), and scheduled daily/weekly/monthly runs. Use for "review all my clients", "which accounts need attention", "what's working across clients", "set up weekly reports", "automate the reviews".

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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. Very specific:

    After every client is saved, run adsgm sweep.

    line 15
  2. Distinctive:

    Learnings change recommendations, never limits: If lessons suggest looser guarded-auto limits, propose it.

    line 29

≈ “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

Portfolio, learning loop and schedules

Load with ads-core. Workflows: portfolio sweep · learning loop · scheduled runs.

Sweep: every client, one view

For each client profile (adsgm clients list):

  1. Read numbers with the platform MCP tools (read-only). Use the last 7 complete days and the prior 7, today and yesterday for spend checks, month-to-date spend, entity rows (campaign / ad set or ad group) with status, learning state, spend, results, CTR (with prior CTR), frequency, daily budget, today's spend and hours elapsed. For Google, also search-term waste and impression share lost to budget.
  2. Build a snapshot (schema) and save it with adsgm snapshot save --file snap.json. Account IDs must belong to that client, or the save is refused.
  3. After every client is saved, run adsgm sweep. It writes portfolio.md, portfolio.html (dashboard) and portfolio.json to ~/.ads-godmode/reports/<date>/ and returns clients ranked RED / AMBER / GREEN, each with its top issues and next step.
  4. Report: the portfolio table first, then per-client issues. Offer to open the clients that need work (/ads:optimize <client>). A sweep never changes anything.

One client at a time for data. Never mix numbers between clients. If a client's accounts can't be read, mark that client "no data" rather than guessing.

Learning loop: get smarter per client

  • When planning a change, attach a prediction: adsgm plan budget ... --predict-metric results --predict-direction up --predict-pct 15 --baseline 11 --baseline-n 11 --read-after 7 --tags budget-up-20. Pick the metric that expresses the goal: budget up → results up; negatives → cpl down; new hook → ctr up.
  • At the read date (or during weekly sweeps), run adsgm learn auto --client X to judge the outcome from the latest snapshot. Or record it by hand with adsgm learn record .... Each outcome is graded worked / failed / inconclusive, with sample-size and noise checks.
  • Creative tests and other insights: adsgm learn lesson --client X --tags ugc-hook --verdict worked --text "...".
  • Before recommending anything for a client, run adsgm learn summary --client X. Lead with what has worked for this client. Don't repeat what failed without a new hypothesis. Label the learnings as CLIENT-SPECIFIC RULE (HIGH or MODERATE when decided on 3+ outcomes, EXPERIMENT below that).
  • Across clients: adsgm learn summary --portfolio shows tactic-level win rates (label: PORTFOLIO HEURISTIC). Use it as a starting hypothesis for a new client, never as evidence about that client.
  • Learnings change recommendations, never limits. If lessons suggest looser guarded-auto limits, propose it. The human edits the profile.

Autopilot (per client, on/off)

Hands-off day-to-day fixes inside budgets the human already approved: pause clear losers, swap tired or failing ads for fixed ones in the same ad set (while the account's total daily budget is at or under the client's cap; max 3 swaps a week), and add small Google negative lists. Never: new campaigns live, budget increases, new geos, tracking, deletes, billing. Switch it with adsgm autopilot on|off --client X --cap <daily max>. Workflow: autopilot. Always mention it during onboarding and ask whether they want it.

Scheduled runs

/ads:schedule sets up daily, weekly and monthly runs. They run with ADS_GODMODE_UNATTENDED=1:

  • Writes are denied unless the client's profile sets automation.allow_unattended_writes: true.
  • Even then, nothing that needs approval can run.
  • Scheduled runs refresh snapshots, judge due predictions, run the sweep and write reports. Problems become recommendations in the report, waiting for you.

Reporting what works / what doesn't

For each client, close every sweep or weekly review with:

  • What worked: from the learning summary and this week's winners, with numbers.
  • What didn't: failed tactics and losing campaigns/ads, with numbers.
  • What to fix next: 3 items max per client, each tagged "needs approval" or "no approval needed", with the exact command or plan.

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.