What's inside
The difference between a useful assistant and an expensive autocomplete is the framing you give it. These prompts constrain the model to the job in front of you.
The 65 prompts cover six areas:
| Area | Prompts | Examples |
|---|---|---|
| Review | 12 | Adversarial diff review, security passes, API contract checks |
| Debug | 11 | Stack trace triage, bisect planning, flaky test investigation |
| Test | 10 | Behaviour-driven cases, edge case discovery, fixture generation |
| Refactor | 9 | Refactor plans, dead code hunts, dependency untangling |
| Design | 12 | Architecture decision records, trade-off analysis, migration plans |
| Communicate | 11 | PR descriptions, incident write-ups, README and API docs |
Language-agnostic
Nothing here assumes a particular stack. Each prompt has a constraints slot where you state the language, framework and anything that cannot change, which is what stops the assistant proposing a rewrite in something you do not use.
What you get
PDF, Markdown and a Notion-ready version. The Markdown file is formatted so you can paste prompts straight into your editor's snippet manager or a CLI assistant.