Sign inGet started
Skills

Anatomy of a finding

View as MarkdownOpen in ChatGPTOpen in Claude
Connect your agent
Add to CursorAdd to VS Code

One finding from start to fix, told as a story: our rule notices wasted ad spend, a skill runs and explains it, and a person decides what to change.

Updated

The other pages describe the pieces. This one follows one problem from the first number to the fix. The shop is made up; how the pieces fit is not.

1. A rule fires

Tuesday morning. Our rule for Wasted ad spend sees a campaign that spent 240 EUR in 7 days with no conversion. Everyone on shop.example who added the skill gets a signal. No model is involved yet.

2. The skill runs

At 06:00 the next day, Wasted ad spend runs on the team owner's Claude. It reads the project's context first, including the team's rule "answer in five lines", then spend by campaign and keyword for the last 7 closed days.

3. The finding

It posts one finding to Maria, who added the skill:

Title: "Brand - broad" spent 240 EUR with no conversion
Summary: The keyword "free analytics tool" took 190 EUR of it. Pause it; keep "nextanalytica".
Numbers: spend 240 EUR, conversions 0, period last 7 days, compared with 31 EUR the week before
Evidence: query spend by keyword, campaign = "Brand - broad"

4. Maria decides

The email "Your agent found something on shop.example" reaches her. She asks her agent "Show me that finding", reads the numbers and pauses the keyword in Google Ads herself. The run didn't touch the ad account, and couldn't have.

5. Next week

She tells her agent "Remember that broad brand keywords didn't convert in November". The next run reads it, and doesn't flag the same thing twice.

From the first number to the paused keyword: one rule, one run, one finding and one decision. The decision was Maria's.

Next step

Run skills on a schedule

Related
  • How a run works: A scheduled run, step by step: it starts, reads context and numbers, reports a finding or stays quiet and finishes.
  • Project knowledge: The team's instructions, goals and memory your agent reads before every answer and every run, who writes them, the limits and how agents add to them.
  • Steer skills: Shape what skills look at and how they report, with team instructions, a skill's own text and feedback.
  • Built-in skills: The built-in analytics skills every nextanalytica project starts with: what each one does, the schedule it gets when added and what it needs.
  • What's in a skill: A skill's parts: a name, when to use it and the instructions, as one SKILL.md you can write yourself.
  • Insights: Signals from our rules and findings from your skills, newest first, with the numbers that prove them.
Still have questions?
Was this page useful?