Code review assistant
Review pull requests for correctness, regressions, security risk, missing tests, and unclear ownership.
Every team reviews code, and agent output is much better when review criteria are reusable.
Starter catalog
Popular AI skills worth publishing first.
Share what already works
Turn daily workflows into reusable packages
If your team already relies on a skill every day — for code review, testing, docs, or onboarding — publish it to AIPM. One package gives every project the same setup, every teammate the same assistant behavior, and your org a simple way to improve and ship updates without copying files by hand. Not sure where to start? Each starter below includes a copy-ready prompt — paste it into Cursor, Claude, or Codex to generate the skill package, then publish it with aipm publish.
Engineering
Test writer
Generate focused unit, integration, and regression tests from changed files, bug reports, or existing behavior.
Testing is one of the highest-repeat development workflows for AI coding agents.
Refactor planner
Ask the assistant to map dependencies, identify safe increments, and produce a verification plan before edits.
Large refactors need planning discipline so AI agents do not make broad, risky changes.
API integration helper
Guide API client changes, schema validation, retries, auth handling, error states, and integration tests.
API work is frequent, repetitive, and easy to get wrong without local conventions.
Frontend
Frontend UX reviewer
Check responsive layouts, accessibility, spacing, text overflow, empty states, and production UI polish.
AI-built UI often needs a reusable product-quality checklist before shipping.
Accessibility checker
Review semantic markup, labels, keyboard flow, color contrast, headings, focus states, and screen-reader hints.
Accessibility rules are concrete enough to package and reuse across projects.
Security
Security audit
Look for leaked secrets, unsafe auth paths, exposed admin routes, injection risks, and risky dependency usage.
Public repos and AI-generated code both need explicit safety checks before release.
QA
Browser test runner
Run local web pages, inspect key flows, capture screenshots, and report visual or interaction problems.
Real browser checks catch layout and interaction bugs that code-only review misses.
Data
Database migration reviewer
Review migrations for locks, rollbacks, indexes, nullability, backfills, and production deployment order.
Database changes are high-risk and benefit from a consistent preflight checklist.
Support
Bug triage and issue summariser
Turn Sentry issues, support tickets, logs, and user reports into impact, likely cause, and next action.
It saves time before an engineer opens the codebase and gives teams consistent handoff notes.
Product
Release notes writer
Summarize commits, PRs, fixed bugs, upgrade notes, known issues, and user-facing changes.
Teams repeat this every release, and the output format can be standardized.
Growth
SEO content reviewer
Review page titles, descriptions, headings, structured data, canonical URLs, internal links, and search intent.
AI products need clear public docs and discoverability from day one.
Knowledge
Repository onboarding
Explain architecture, folders, local setup, test commands, release flow, and common development traps.
New contributors and AI agents both need durable project context.
Documentation maintainer
Keep READMEs, changelogs, runbooks, examples, and onboarding docs aligned with code changes.
Documentation drift is constant, and skills make the expected style repeatable.
Tools
MCP server setup
Install and document Model Context Protocol servers, prompts, resources, tool permissions, and verification steps.
MCP is becoming the common way to connect AI assistants to tools and context.
Operations
Spreadsheet analyst
Analyze CSV/XLSX files, normalize columns, build summaries, create formulas, and prepare chart-ready outputs.
Teams use AI for lightweight analysis long before they build a formal data pipeline.
Presentation builder
Turn notes, specs, or reports into crisp slide outlines with audience, narrative, and visual QA guidance.
Reusable presentation workflows help agents produce structured decks instead of loose summaries.
Creative
Image prompt designer
Create visual prompts, style constraints, asset briefs, and review checklists for generated images.
Design and product teams need repeatable image direction, not one-off prompt guessing.
How we chose this list
The list focuses on repeated workflows that AI coding tools and assistants already support well: reusable project rules, codebase understanding, testing, Git operations, browser checks, documentation, MCP setup, and structured knowledge work.
The goal is to seed AIPM with useful public packages while respecting other projects. Before a starter becomes a live registry package, it should have original content, clear ownership, a license, and target-specific files that install cleanly.
Reference signals
Cursor RulesProject rules are reusable, version-controlled instructions for AI coding workflows.Claude Code common workflowsCommon terminal-agent workflows include understanding code, editing, testing, and Git operations.OpenAI Codex use casesCodex workflows cover production systems, web development, native development, collaboration, and saved skills.Model Context ProtocolMCP servers expose tools, resources, and prompts that AI applications can discover and use.