---
name: creative-fatigue-signal-scanner
description: "Scan a paid social account (Meta, TikTok, or similar) for creative fatigue. Flags decaying ads only on two or more combined signals (frequency up + CTR down, cost per result up), ranked by severity — detection only, read-only, no creative suggestions. Use for weekly fatigue checks or when cost per result trends up without an obvious cause."
metadata:
  version: "1.1.0"
  scope: "standalone-template"
---

> Standalone draft/audit template. Not validated for OLLIE execution.


# Creative Fatigue Signal Scanner

## Who it's for

Agencies and in-house teams running paid social (Meta, TikTok, or similar) for clients where creative gets stale faster than anyone has time to notice — which is most accounts running more than 2-3 active creatives at once.

## What problem it solves

Creative fatigue is usually caught late — after cost per result has already crept up for a week or two and a client has started asking questions. This skill gives an agent the job of watching the signals continuously and telling you exactly which creative is decaying and why, so a human can decide what to do about it before the client notices.

## When to use it

- Weekly, as a standing check on any account with live paid social creative
- Whenever cost per result on an account trends up without an obvious external cause (seasonality, targeting change, etc.)
- Before a monthly client review, so you walk in already knowing what needs a refresh

## Required inputs

- Read-only access to the ad platform's performance data (spend, frequency, CTR, cost per result) for each active ad, over a defined lookback window
- A simple list of which ads are currently live and how long each has been running
- Optional: the account's historical average CTR/cost-per-result, so "fatigue" is judged against this account's own baseline, not a generic industry number

## Agent instructions

You are scanning a paid social account for creative fatigue. You have
READ-ONLY access to performance data. You may not pause, edit, duplicate, or
touch any live ad, and you may not generate or suggest replacement creative —
this task is detection only.

1. Pull performance data for every currently active ad: spend, frequency,
   CTR, and cost per result, over the specified lookback window.
2. For each ad, compare current-period performance against its own
   first-two-weeks-live baseline (or the account's historical average if the
   ad is too new to have its own baseline).
3. Flag an ad as a fatigue signal if you see two or more of the following
   together, not just one in isolation:
   - Frequency rising while CTR falls
   - Cost per result rising while spend holds flat or rises
   - A meaningful CTR decline that isn't explained by an audience or
     targeting change you're aware of
4. For each flagged ad, note how many days it's been live and rank all
   flagged ads by severity (how far off its own baseline it has drifted).
5. Do not recommend specific new creative concepts, do not touch the ad in
   any way, and do not contact anyone about it. Produce the findings list and
   stop.
6. If performance data is too sparse to judge fatigue confidently (e.g., an
   ad has been live less than a week), say so explicitly rather than forcing
   a verdict.

## Expected output format

A ranked list: Ad Name/ID | Days Live | Fatigue Signal(s) Detected | Severity (High/Medium/Low) | One-line "why this matters."

Followed by a short summary: how many ads were reviewed, how many were flagged, and which single ad most urgently needs a human look this week.

## Human approval checkpoint

Nothing about the creative or the account changes based on this scan. A human (media buyer, campaign manager, or the client) decides what to do with each flagged ad — pause it, brief a refresh, or leave it running with a documented reason.

## Safety rules

- Read-only against ad performance data, full stop
- Fatigue is only flagged with two or more supporting signals together, never on a single noisy metric
- Sparse data leads to "not enough data yet," never a forced verdict

## What the agent must not do

- Pause, duplicate, or edit any live ad
- Generate or suggest replacement creative concepts (that's a separate, later step with its own approval gate)
- Contact a client, designer, or account owner directly — it produces a report for a human to act on

---

Part of the Outloop Agency Skill Pack ([outloop.co](https://outloop.co)). To turn this from a draft-only workflow into a safe runtime workflow across client workspaces, use Outloop to approve the right access, block wrong-client use, and keep secrets out of the agent context.