Honest weekly digest

What it does

A weekly report is a trust instrument. The moment it states one number the reader can falsify from their own Ads Manager or CRM, every future report is discounted.

This skill builds the digest and the guards that keep it honest. The guards are the point: any tool can compute week-over-week deltas; knowing when not to state them is the skill. The digest covers per-platform spend, results and cost per result with WoW deltas, the top-spending campaigns, pipeline movement if you provide CRM stage history, and a one-line reconciliation callout of platform claims vs CRM-recorded results.

Quick start

Terminal

git clone https://github.com/PierreNicolasMarquez/pierce-skills.git
cp -r pierce-skills/skills/honest-weekly-digest ~/.claude/skills/

Then in Claude Code: "build my honest weekly digest from these exports".

The full SKILL.md, with the exact API calls, lives in the pierce-skills repo. MIT-licensed, runs on your token, never phones home.

How it works

Six guards apply to every digest, all of them, in this order. A comparison only ships when every guard lets it through.

1
G1: Short-history suppression
If stage or tracking history covers less than two complete weeks, WoW comparisons on that data are suppressed entirely, replaced by a tracked-since date and the date comparisons start. Never compare a full week against a partial one.
2
G2: Cleanup-burst detection
The skill scans stage history for bursts: many stage changes in a short window, like tens of transitions inside an hour or one operator restaging a whole pipeline in a day. That is CRM housekeeping, not funnel movement. When detected, the movement section carries a can-include-cleanup-restages note and any headline claim built on the burst is withheld. A stock like 92 clients on the book stays honest where a flow like 29 deals won this week, inflated by the burst, does not.
3
G3: Small-number suppression
No percentages or cost-per-result on denominators too small to mean anything: under about 5 results or a handful of euros of spend gets raw counts instead. A week with 1 lead does not have a CPL trend.
4
G4: Freshness guard
Before reporting 0 spend or 0 results on any platform, the skill verifies the data is fresh: does the API return rows for yesterday? A dead sync reads exactly like a dead account. If freshness is unverifiable, the digest says no data received, never 0.
5
G5: Thin-history footnote
When ad history on a platform starts mid-window, that platform's row carries its own start-date note rather than silently comparing unequal periods.
6
G6: Caveats retire themselves
Every guard states when it stops applying, for example the date WoW comparisons become clean. A permanent caveat is an unfixed bug.

Caveats

Output style is plain language, real numbers, no adjectives doing a number's job. Every claim traceable to a source the reader can open.
If the honest digest is boring, ship the boring digest.