th-auditlisted

# th-audit The only honest source of what works for an account is that account. Every rule in this pack is a starting guess. The user's own last 30 posts are the evidence. ## Input Ask for whichever the user has: - Screenshots of the insights for each post (views, likes, replies, reposts, quotes, shares), plus their follower count. - Or the posts with their numbers typed out. - Or just the posts and their like and reply counts, which is enough for a first pass. Also read `~/.claude/threads/log.md` if it exists. It records the format and the hook formula each post used. ## What to measure Views mostly track how many people already follow the account, so on their own they are the least useful number. Compute these, and show the working: | metric | how | what it tells you | | --- | --- | --- | | **Reply rate** | replies / views | whether the post started a conversation. The main signal on Threads | | **Engagement rate** | (likes + replies + reposts + quotes) / views | whether the post earned the views it got | | **Reply ratio** | replies / likes | whether people talked, or only nodded | | **Reach multiple** | views / followers | whether it travelled past existing followers | | **Repost + quote rate** | (reposts + quotes) / views | whether people wanted to be seen sharing it | Rank by reply rate and reach multiple, not by views. A post with 800 views and 30 replies beat the one with 9,000 views and 4. ## Then find the pattern Put the top 5 and bottom 5 side by side