REPO PACKAGE
mastra-ai mastra
COMMUNITY31skills
21ready to use
LABSCO SUMMARY
The engineering skills are documentation, testing, and review standards this repository enforces on itself: code-standards, security-review, and performance-review work as PR checklists; playground-msw-tests, e2e-tests-studio, testing-core-processors, builder-smoke-test, mastra-smoke-test, and smoke-test form a layered test strategy for the Studio UI and CLI; mastra-docs and docs-audit cover the documentation site; mastra-frontend, react-best-practices, and tailwind-v4 cover its own design system; and pr-explainer, pr-splitter, debugging-difficult-bugs, general-tasks, and research-tasks are day-to-day process guides for whoever, human or agent, is working in this codebase. The other eight are a different thing entirely: agent-prompt-quality-bar is a shared final-review rubric, and coding-agent, content-writer-agent, customer-support-agent, generic-assistant, ops-automation-agent, research-agent, and spreadsheet-agent are one playbook each for drafting the system prompt of a Mastra-built agent of that type.
That split means two different readers get value here. Someone opening a pull request against mastra-ai/mastra gets the engineering skills as a repo-specific style guide; someone building their own agent on the Mastra framework — a TypeScript framework for agents, workflows, and MCP servers, per its own README — gets the eight archetype playbooks as a starting point for a system prompt, and nothing else in the package is aimed at them. 21 of the 31 need nothing beyond the skill file; five assume a local tool, three a stored account key, and two an MCP connection, Chrome MCP among them for the browser-driven Studio check in smoke-test.
READ THE FULL ANALYSIS
Two of the 31 fit neither group, and we can't explain them from what we have. severe-weather's entire description is "Use when conditions include storms, flooding, heat waves, or other hazards" — nothing ties it to Mastra, agents, or this repository. ralph-plan builds "ralph-loop commands," a term that appears nowhere else in our material, so we're naming it without saying what it does.
One of the 31 has no description at all. We have material for 30 of them; the last is live in the repository but frontmatter-empty, so it isn't named above and we are not guessing at its purpose from its filename.
No skill in this package carries its own license tag, though the repository as a whole does. The README describes a dual-license model — Apache-2.0 for the core framework, a separate Mastra Enterprise License for anything under an ee/ directory — but that split isn't recorded skill by skill in our data, so check which side of that line a given skill's own folder sits on before reusing it commercially.
Checked 19 September 2026 from the repository's full skill catalog and its README.
ALSO IN THIS PACKAGE
PROMPT TEMPLATES
10 itemsPROMPT TEMPLATE
changeset.prompt
Create a changeset for version bumps
PROMPT TEMPLATE
commit.prompt
Commit the work you and the user have been doing
PROMPT TEMPLATE
gh bulk issues.prompt
PROMPT TEMPLATE
gh debug issue.prompt
Debug a GitHub issue
PROMPT TEMPLATE
gh fix ci.prompt
Use the GH CLI to fix CI failures on a pull request
PROMPT TEMPLATE
gh new pr.prompt
Create a new pull request on GitHub
PROMPT TEMPLATE
gh pr comments.prompt
Handle PR review comments on the current branch's pull request
PROMPT TEMPLATE
make moves.prompt
Make moves on a GitHub issue
PROMPT TEMPLATE
pr.prompt
Create a changeset and open a pull request
PROMPT TEMPLATE
ralph plan.prompt
Interactive planning assistant that helps create focused, well-structured ralph-loop commands through collaborative conversation
WHAT'S INSIDE
31 showing · 31 totalReady to use
Nothing else to set up — install it and go.
31 of 31
SETUP
Showing all 31 skills
agent-prompt-quality-bar
The standard an AI agent's instructions have to meet before it goes live: one clear job, a definition of finished, and an honest answer when it cannot do something.
builder-smoke-test
Puts Mastra's Agent Builder through a full test run on a throwaway project and reports, section by section, what passed and what broke.
code-standards
Reviews code changes in a fixed order — real bugs first, then quality, then house style — and separates the must-fix items from the nice-to-haves.
codemod-patterns
Instead of updating outdated Mastra code by hand across hundreds of files, you build a small script that does the rewrite — test cases first, script second.
coding-agent
Spells out the instructions for an AI that writes and fixes code: what it may edit, how it checks its own work, and when it is allowed to call the job done.
content-writer-agent
Sets up an AI writer that turns a topic into a finished draft — blog post, newsletter, LinkedIn post — instead of a brainstorm you still have to write up yourself.
customer-support-agent
The blueprint for a support agent that reads each incoming message, drafts the reply, and knows which cases a human has to take instead.
debugging-difficult-bugs
A rule for hard bugs: prove what the code actually did before you change a line, instead of shipping a second guess and hoping.
docs-audit
Checks whether Mastra's docs still describe the code accurately, and reports each mismatch with the source line that proves it.
e2e-tests-studio
Browser tests for Mastra's Studio that prove a feature actually does its job, instead of only checking that a button or a spinner showed up.
general-tasks
A default way of working through a vague request: state the goal, list the steps before taking them, then take them one at a time.
generic-assistant
The fallback recipe for a personal helper that fits no specialised category — it still has to name the jobs it handles and turn down everything else.
mastra-docs
Documentation guidelines for the Mastra project: they say which guide governs each kind of page, how to word and lay pages out, and what to do when one is moved, renamed, or deleted.
mastra-frontend
The line between look and layout in Mastra's own interface code: colours, type and shadows come from the design system, and you only place things on the page.
mastra-smoke-test
Exercises a real Mastra project, on your machine or against a live deployment, and reports which parts still work and which ones broke.
ops-automation-agent
Sets up an agent for the jobs that repeat — a daily post, a sync between two systems, an alert — and makes it prove afterwards what it actually did.
performance-review
A reviewer's checklist for the things that quietly make software slow, covering the database, memory use, what the screen redraws, network calls and slow calculations.
playground-msw-tests
Test Mastra's playground interface by letting the real code run and faking only the server's replies, so a test breaks the moment the app and the server stop agreeing about the data.
pr-explainer
Turns a pile of code changes into a single web page that explains them to whoever has to review it: what the problem was, what happens differently now, and what was run to prove it works.
pr-splitter
A change too big for anyone to review, taken apart into a series of small ones that each stand on their own, with a running note of what moved where so nothing is quietly lost.
ralph-plan
Asks you questions until a vague idea has become a written brief an AI coding agent can work through unattended: the background, what to set up first, the tasks in order, and how to tell it worked.
react-best-practices
Twenty-six house rules for React code from the Mastra team, ordered so the ones that most affect how quickly a page appears come first.
research-agent
Writes the instructions for an AI researcher of your own — one that reads up on a topic you name and comes back with a short brief where every claim says which source it came from.
research-tasks
Answers a question by going and looking it up on the web, and gives back a single paragraph with the list of sources it actually read underneath.
security-review
A pull request read the way an attacker would read it, with each thing found ranked from critical down to a minor suggestion.
severe-weather
When a forecast involves a hazard like a storm, flood, extreme heat or cold, or high wind, you get a one-line safety note ahead of the weather details.
smoke-test
Creates a brand-new Mastra project from the starter, then drives its Studio around in Chrome — pages, chat, a tool run — and reports what held up.
spreadsheet-agent
Builds the agent that reads and updates your spreadsheets — Google Sheets, Excel, Airtable, Notion — for the rows-and-columns work nobody wants to do by hand.
tailwind-v4
Tailwind's version 4 renamed classes and moved configuration into the stylesheet — this is the current way to write it, and the old habits to stop shipping.
testing-core-processors
Tests Mastra's error handling against a fake model that fails on command, so a retry path can be proved without waiting for a real outage.
testing-mastracode-tui
Builds mastracode's terminal app and drives it by hand through the things that only break in a real terminal: model setup, thread state, and how the screen draws.
No skill in this package matches that.
HOW TO GET IT
WHOLE PACKAGEInstall the whole package (31 skills)
npx skills add mastra-ai/mastraSINGLE SKILLOr install a single skill
npx skills add mastra-ai/mastra --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.