Substack stats

Skill gokul-kulkarni/content-studio-plugin/skills/substack-stats

Turn Substack analytics into the normalized record shape the analyze skill reads — subscribers (free + paid), open rate, subscriber growth, and per-post views, opens, clicks, reactions, and comments. Works from the exports you download from Substack (the full ZIP export, the Dashboard Stats export, or a pasted table), with an optional public-endpoint fetch for reactions/comments. Use whenever someone asks how their Substack is doing, wants post-level or subscriber numbers, asks 'which essay performed best', or is gathering data for an outreach/quality analysis.From its SKILL.md

Install

npx -y skills add gokul-kulkarni/content-studio-plugin --skill substack-stats

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.5 KB, as published. Nobody here has run it

Substack Stats

Turn the numbers for $ARGUMENTS (a Substack publication) into the normalized record shape that analyze reads. You bring the data — download it from Substack and drop it in; this skill parses it. In rough order of richness:

Best source — the full export ZIP

Settings → Export produces a ZIP. It's the richest and most honest source:

posts.csv — every post: title, subtitle, audience (everyone/only_paid), type, send times.
email_list.gokulkulkarni.csv — every subscriber with created_at, plan, email_disabled, first_payment_at. Map each created_at to the post that ran that day to see which posts convert.
posts/NNN.delivers.csv / posts/NNN.opens.csv — per-post subscriber-level delivery and unique opens, with device/geo. This is the only source of true unique-subscriber opens (vs. the inflated dashboard).

Parse the ZIP directly (unzip, read the CSVs). This gives subscriber growth, per-post reach/opens, loyal-reader concentration, and list health (paid, unsubscribes) — everything except web view counts.

Also useful — the Dashboard Stats export or a paste

Dashboard → Stats → export/download gives per-post views and open/click rates; the Subscribers tab exports counts and growth. A pasted or screenshotted Stats table works too. Use this to fill web views that the ZIP doesn't carry.

Optional — public inventory (no download, no auth)

If there's no export handy, the public JSON endpoint gives reactions/comments/inventory. WebFetch:

https://<publication>.substack.com/api/v1/posts?limit=50&offset=0

Per post: title, post_date, reaction_count, comment_count, audience, canonical_url, type. This is likes/comments/timing/free-vs-paid only — no views, opens, or subscriber counts (never public). Use the publication subdomain (e.g. yourname.substack.com), not the reader profile substack.com/@handle.

⚠️ Open rate is inflated by Gmail's image proxy. Substack's "opens"/"open_rate" count image-proxy prefetches (Google opens the tracking pixel for the reader, logged as country US regardless of location). Treat open rate as a directional title signal, not an absolute — for the real number use opens.csv (unique subscriber opens) from the full export, and don't read US geo as real location. Pass this caveat to analyze with any open-rate figure.

What to produce

Emit the normalized records (schema in the analyze skill): one PostRecord per post plus one AccountSummary. Fill what you have, leave the rest null, and note the source (export / public) so analyze can weight confidence. Then hand off to analyze — this skill gathers; analyze interprets.

Privacy

These are personal analytics. Keep them local (the working dir or scratchpad); never post them to an external service.