fb-audit

The user's own numbers beat any advice in this pack. This skill reads them.

Collect

From Meta Business Suite > Insights > Content (or Professional Dashboard), the last 30-90 days of posts. An export is best; screenshots work. Per post: date, format, first line, reach, views, reactions, comments, shares, link clicks, messages started (if shown), follows from post.

Also read ~/.claude/facebook/log.md for which hook formula each post used.

Rank on the right things

1. Outlier multiple: each post's reach divided by the median reach of the posts in this window. Same logic as ../fb-viral/swipe.py (you can feed it a TSV with account set to the Page name and metric = reach).
2. Shares per 1,000 reach: the strongest distribution signal on Facebook.
3. Messages / clicks per 1,000 reach: the business signal.
4. Views alone are last. A post that got 50k views and no messages may be a worse post than one with 3k reach and 12 enquiries.

Report

AUDIT · 42 posts · Aug 1 - Oct 27 · median reach 1,840
TOP BY MULTIPLE
  6.2x  Reel   "If you live in Patan..."      shares/1k 14.1  msgs 9
  4.8x  Photo  6am prep, 4 photos             shares/1k 8.0   msgs 3
BOTTOM
  0.2x  Link   "New blog post is up!"         shares/1k 0.4   msgs 0
PATTERNS
  - Local-named hooks: 4 of top 5
  - Link posts: median 0.3x (9 posts)
  - Shared-from-other-Pages: median 0.2x
RULES FOR NEXT MONTH
  1. Two Local Angle Reels a week.
  2. Links go in first comments only.
  3. Stop sharing other Pages' content.

Rules must come from the data. If the sample is under ~15 posts, say the patterns are weak. Name confounders (a holiday, a boosted post, a Group share).

Honest limits

Boosted/paid posts are separate; exclude or mark them. Facebook changes what Insights shows; use what's there, don't fill gaps with guesses. Never invent a number. Feed the rules to /fb-plan.