Long-Form Editor — Structural Pass
Edits a 3,000-word draft for structure before style: argument order, missing evidence, dead sections. Returns a diff plan.
Reviews a diff for correctness, then security, then readability — in that order, with severity tags and a fix per finding.
Generated with this exact template on claude. No retouching.
You are a staff engineer reviewing a diff. Review in order: correctness, security, readability. Tag each finding P0/P1/P2, cite the line, state the failure mode, and give a concrete fix. Do not comment on style nits.
The full template is 4× longer than this preview, and includes every variable with its tested example.
Licenza: uso commerciale dell'output, rivendita del testo vietata.
| Variable | What to put in it | Example used |
|---|---|---|
| diff | Paste the diff | (git diff output) |
| context | What the code is for | (one or two sentences) |
AI code review tends to produce a flat list of thirty observations, which is worse than no review because it buries the one P0. This prompt splits the review into three ordered passes and forbids style commentary entirely, so the output is short enough to act on.
The failure-mode requirement is the part that changes behaviour. "This could throw" is not reviewable; "throws when `items` is empty because reduce has no initial value" is. Asking for the smallest fix as a snippet means the reviewer has to actually solve it rather than gesture at it, and the merge verdict forces a decision instead of a hedge.
Example output
Generated with this template — full output shown on the listing.
Qui Tanaka @quietcode
Staff engineer. Code review, SQL, research synthesis, triage.
The full claude template with all 2 variables, the example values we tested with, and permanent access from your dashboard. Outputs you generate are yours to use commercially.
Yes. You get commercial rights to anything the prompt produces. You may not resell or republish the prompt text itself, or pass it off as your own listing.
Request a refund within 24 hours of purchase. We compare the listing against your result — if the prompt does not do what this page claims, you get your money back.
Edits a 3,000-word draft for structure before style: argument order, missing evidence, dead sections. Returns a diff plan.
Takes a slow query plus schema and returns a rewrite, the index it needs, and the plan change to expect.
Turns a messy process into an SOP with preconditions, decision points, rollback, and an owner per step.
Sorts a batch of emails into reply-now, reply-later, delegate, archive — with the reply drafted for the first group.