Review

Once a draft exists (see Production), review makes it stronger and then stretches it further. Two movements: refine the piece you have, then repurpose it into more.

Refine

These three are complementary — they do different jobs, and they work well in sequence.

1. diagnose — find what’s wrong. Flags structural issues against the post-structure playbook: slow intros, blurring main points, low Rate of Revelation, forced conclusions, headline/intro mismatch, rhythm problems. It only flags and suggests; it does not rewrite. Run it first to know where to focus.
2. rate — see where it stands. A structured quality read: VPM (value per minute) plus scores across the quality dimensions. Use it to gauge readiness (“is this ready to ship?”) or to compare two versions. It only evaluates.
3. improve-writing — fix it. Takes a draft that already has structure and tightens, restructures, or finishes it. This is the skill that actually changes the text — run it after diagnosis and rating, pointing it at the issues they surfaced.

A typical loop: diagnose → rate → improve-writing, repeating until the rating clears your bar.

Diagnose this draft and tell me why it doesn't land yet.
Rate it so I know where it stands.
Now improve it, focusing on the issues you found.

Guidance, not ground truth

rate and diagnose reflect the framework’s judgment, not an objective measure of quality. Use them to focus your attention — then make the call yourself. See Known Limitations.

Rating LLMs has limits

Rating is context-sensitive: an LLM tends to rate more favorably the more it’s used within one session (it mimics a human feedback loop), so run rate once per fresh session/context for a clean read. A piece scoring above ~80 is usually fine — beyond that an LLM will keep finding nitpicks and escalating them into “issues.” Treat the score as a signal, not a verdict.

Repurpose

Once a piece is strong, get more out of it.

• tldr — a one-paragraph abstract (~100–150 words) for the top of a long article, in your voice, so a reader can decide whether to commit.
• distill — break a long piece into shorter posts: Micros (≤280 chars for X / Bluesky / Threads / Mastodon) or LinkedIn shorts (800–1,200 chars), or both. Ideal for cross-posting.

Write a TLDR for the top of this article.
Distill this piece into five X micros and two LinkedIn shorts.

Illustrate

Experimental

illustrate is experimental — it targets Midjourney only and the end-to-end image workflow is just partially built.

Run illustrate on a publish-ready piece to produce Midjourney prompts for companion imagery — banner, header, hero, section illustration, or social share graphic.

When it’s done

The piece clears your rate bar, reads the way diagnose and improve-writing shaped it, and has the abstract, repurposed posts, and imagery you need. For the per-skill reference, see the skills index.