li-auditlisted
Install: claude install-skill reportedguy/li-stack
# li-audit
The only honest source of what works for an account is that account. Every
rule in this bundle, `../../shared/algorithm.md` included, is a prior. The user's own last
20-30 posts are the evidence.
## Input
Whichever the user has:
- The analytics export (LinkedIn: Analytics → Content → Export), CSV or XLSX.
- A screenshot per post with impressions, reactions, comments, reposts.
- Just the posts and their reaction counts (enough for a first pass).
- The engagers summary from `/li-engagers`, if it ran.
Also read `~/.claude/linkedin/log.md` for the hook and angle each post used,
and the follower count at the time.
Work the numbers in code (a short Python script; pandas and openpyxl read
the export), not in your head, and show the working. On Windows run
`python`, not `python3`.
## Measure these, not impressions
Raw impressions mostly reflect how many people already follow you.
| Metric | How | What it tells you |
|---|---|---|
| **Engagement rate** | (reactions + comments + reposts) / impressions | did the post earn its reach |
| **Comment ratio** | comments / reactions | did it start something or just get a nod |
| **Reach multiple** | impressions / follower count | did it travel past your existing audience |
| **Save / send rate** | if available | the strongest predictor of future reach |
Rank by engagement rate and reach multiple. A post with 900 impressions and
40 comments beat the one with 12,000 impressions and 6.
Compare against the follower-count band