Tag
A lot of repetitive work can be handed off with the right prompt. This set covers breaking tasks into steps, documenting processes, and keeping projects moving without extra manual effort.
19 prompts · page 2 of 3
Describe the issue you've spotted in plain English along with the specific lines and fuller details, and get back a ready-to-paste GitHub review comment instead of staring at a blank comment box. The output is specific, explains why the issue matters, includes a concrete suggested fix, and skips filler openers like "Great work, but..." — also useful as a template for junior reviewers learning to write good comments.
Paste a diff to get a review that stays narrowly focused on logic errors, null dereferences, race conditions, and off-by-one bugs — the kind that actually reach production — while ignoring formatting and naming entirely. Each bug found comes with the exact line, why it's a bug, and a suggested fix, and it says so explicitly if nothing turns up rather than padding the review with style notes.
Takes review notes you've already worked out — the issues, not the raw diff — and turns them into 5 polished GitHub PR comments ready to post. Each one stays direct without being harsh, names the specific issue, explains its impact, and suggests a fix, capped at 80 words so reviewers actually read it. Useful when you already know what's wrong and just want the wording tightened before publishing.
A fixed, project-specific checklist for reviewing changes in a React/TypeScript plus FastAPI/Python codebase, not a generic "review this code" prompt. - Frontend: React Flow node patterns, Zustand store selectors, design tokens - Backend: Pydantic model layering, FastAPI auth dependencies, response models - Also covers security, performance, test coverage, docs, and commit hygiene - Reach for it before merging, to keep reviews consistent across the team
A reusable context block to paste at the start of any AI coding session, covering your stack, key files, conventions, and known constraints so replies stop being generic. - Fill in project name, stack, main/config files, and naming/testing conventions once - Lists known constraints like performance limits and technical debt to work around - Save it as a markdown file and refresh it weekly as the project changes - Re-paste key facts mid-session on long threads to keep the AI anchored
Turns a pasted directory tree into a guided tour of an unfamiliar codebase, aimed at whoever is starting fresh or inheriting legacy code. - Identifies the architecture pattern, entry points, and the 5 most important files - Traces data flow, external dependencies, and maintenance red flags - Points to exactly which files to open first for a stated task - A follow-up step explains any single pasted file in depth: role, dependents, gotchas
Fill in the feature you want built, the language, and up to three specific requirements, and get an implementation request structured enough that nothing important gets skipped. It explicitly asks for error handling, edge cases, performance, and framework best practices to be considered up front, and instructs the output to keep existing comments intact and explain its own logic inline.
Reviews a pasted PR or diff in two passes instead of one: the first pass lists every potential issue it can find, and the second pass critiques those findings and throws out anything weak or unsupported. The final output separates confirmed issues from possible risks and lists discarded findings with the reason they didn't hold up, which cuts down on the false alarms a single-pass review tends to produce.
Paste a draft agent skill or reusable prompt along with your project context, the target task, available tools, and known risks, and get back a structured audit covering purpose, required inputs, output format, safety boundaries, and failure modes. The review also flags missing checks and repeatability gaps, then returns an improved version ready to test — useful before wiring a skill into an autonomous coding agent or handing it to a team.