Checks a live web page for barriers like keyboard traps, low contrast, missing labels and focus problems, with a fix for each.
androoAGI/starnet
A collection of ready-made skills that let a local AI agent crew handle real work like coding, writing, research, marketing and everyday admin.
- 102 skills
- 1.6k stars
The 5 most installed
Writes paid-ad variants that each test one variable, based on ads already running, and reports honestly which won.
Works through pull request review comments, makes the right code changes, answers the rest, and holds replies for your approval.
Tries to break the work and then refute its own findings before reporting anything.
Turns a product launch into channel-shaped copy that leads with the hook and keeps the facts right.
The other 97 skills
Reads a job posting and drafts an application that maps your real history onto its language.
Keeps a list of recurring charges with costs, renewal dates and cancel steps, flagging price rises, duplicates and expiring trials.
Recovers a dated archive copy or the site's own data feed when a page is blocked, rate-limited or gone, and cites it honestly.
Makes images, covers and transparent cutouts that share one pinned style, with every file checked against it.
Drives a real website step by step, reading the live page before each click and stopping at anything irreversible.
Finds meeting times that fit your calendar, hours and time zones, offers three ranked options and drafts the invite without booking.
Writes a changelog between two versions, grouped by what users notice, with breaking changes and fixes first.
Sends short, phone-friendly updates to Telegram, Discord, Slack, Matrix or Signal, and knows when not to message.
Generates a reference wiki for a codebase with architecture overview, module deep dives and Mermaid diagrams, based on files it actually read.
Drafts first-contact business emails and follow-ups that sound human, with every personal detail taken from a source actually read.
Tracks what you said you would do, notes honest slippage and suggests one next action.
Watches named companies and sends a short cited digest only when something matters, staying quiet otherwise.
Keeps a living publish calendar of what ships where and when, adapted per platform and checked before staging.
Reads a contract, terms, lease or policy closely and surfaces the clauses with real exposure, quoted and ranked by cost.
Reviews a landing page or form on desktop and phone to find why visitors don't sign up, and ranks the fixes.
Sweeps real spending line by line, checks cheaper alternatives against live prices and ranks the savings.
Hands a job to another agent with a brief it can finish alone, including context, a checkable definition of done and proof of return.
Mines reviews, forums and call transcripts for what customers really want, quoting a source for every insight.
Sends one short morning message with today's calendar, mail needing replies, open tasks, weather and one thing to watch.
Turns a messy spreadsheet or CSV into a clean table, plain-English findings and a chart, recording every cleaning decision.
Turns scattered live web pages into one structured dataset where every row traces back to its source page.
Turns a decision you can't make alone into a focused questionnaire for the person who has the missing knowledge.
Finishes work as openable files with a clear name and place, and points you to the main one.
Gets an app from working locally to actually live, fixing the real build error and verifying the deployed site responds.
Composes a recurring digest that gathers the period's items, distills the signal and cites every claim.
Extracts facts, obligations, deadlines and tasks from contracts and reports, keeping page citations and how binding each item is.
Writes an email sequence where each send has one job, testable subject lines and a clear reason to exist.
Hunts the ways an AI-built app leaks, such as keys shipped to the browser, open database rules and public buckets.
Finds what is publicly exposed about you, ranks it by what it actually enables, and gives exact removal steps.
Watches a source for changes against a baseline and alerts only when it crosses the bar you set.
Inventories, sorts, renames and de-duplicates your files without ever destroying data.
Wraps up finished code work safely, confirming tests pass and letting you choose to merge, open a pull request or keep it.
Builds a weekly training and eating plan for your goal, logs sessions and weight, and adjusts from the log.
Finds why automated checks failed, reproduces the failure locally, fixes the real cause and proves it passes.
Turns your notes into question-and-answer cards, quizzes you with spaced repetition and grades you honestly.
Gives a web page or app its own look by choosing colors, type and layout on purpose, then critiquing the result.
Stress-tests a plan before code is written by listing its key decisions and questioning you on each with a recommendation.
Answers a question so every claim carries a numbered source and the exact sentence backing it up.
Drafts the message you've been avoiding, such as a boundary, apology or bad news, clear and kind.
Organizes health records, tracks symptoms factually and builds questions for an appointment, without diagnosing or dosing.
Writes the role honestly, screens candidates against criteria set in advance and interviews for evidence rather than impressions.
Sorts every inbox into one board from urgent to noise, with draft replies held and never auto-sent.
Sorts a backlog of issues and pull requests, labels by type and severity, links duplicates and turns accepted ones into briefs.
Turns a trip or outing into a verified day-by-day itinerary with real prices and times.
Writes a landing page for one audience with one promise up top and objections answered in order.
Plans a free resource people trade their email for, from topic and outline to promotion and lead measurement.
Finds where ideal customers gather and builds evidence-backed lead lists from public sources.
Logs ledger entries and reconciles totals with computed math instead of eyeballing or forcing a balance.
Designs a repeatable workflow as stages with named handoffs, briefs and a test before it goes live.
Turns an article, PDF, thread or captioned video into a gist, sourced key points and quotes.
Builds and maintains an interlinked Markdown knowledge base that flags contradictions and answers questions from it.
Answers location questions with free OpenStreetMap data: geocoding, nearby places, distances, directions and timezones.
Builds a channel and campaign plan grounded in live audience research, with a measurable goal per play.
Builds and tests a Model Context Protocol server so AI agents can use an outside service through designed tools.
Plans a week of meals around what you already have, actually cook, and can manage on a weeknight.
Turns meeting notes into cited decisions, owned action items and a follow-up draft for your approval.
Writes a one-page brief before a meeting: attendees, goal, agenda, open items, questions and an opening line.
Builds an evidence case for a better price, refund or rate, then drafts the opening, rebuttals and walk-away.
Cleans the agent's notebook by finding duplicates and contradictions, correcting stale beliefs and listing what to delete.
Reads, searches, creates and edits notes in your local Markdown vault, linking them with wikilinks.
Finds money-making openings that fit your skills and resources, sized with live demand evidence and ranked.
Keeps reminding you about bills, refills and chores until you mark them done, with snooze and a weekly summary.
Builds the story a pitch deck must carry, with every number sourced and the weakest slide named.
Fetches live like-for-like prices, compares the true total and tells you whether to buy or wait.
Works out what to charge and how to package it, including tiers, first price and when to raise it.
Learns a new project folder before touching it: house rules, layout, build and test steps, saved as a sourced map.
Puts a finished website online on your own GitHub Pages, Cloudflare Pages or Netlify account with a rollback path.
Turns a pile of receipts and invoices into a clean ledger and folder plan, with every figure traced to its file.
Handles code review feedback carefully: checks each point against the real code, fixes what is right and pushes back with reasons.
Keeps a durable record of the people you deal with: what matters to them, promises made and what is owed next.
Traces REST and GraphQL API failures layer by layer and gives a clean reproduction for each finding.
Writes an unattended routine with a schedule, stop rules, no duplicates and a known place for results.
Designs a data model and access rules that keep each user's data separate.
Scans a codebase for secrets, injection and missing permission checks, ranking each finding with proof.
Checks a website for search visibility problems and ranks each finding by impact with the page cited.
Writes timed, beat-marked short video scripts built for retention in a spoken voice.
Rewrites or audits technical text using Simplified Technical English rules for docs, runbooks and error messages.
Writes song structure, rhymes, hooks and lyrics, plus style prompts for AI music tools like Suno.
Turns how someone actually does a task into a procedure another person can follow without questions.
Confirms an important claim against several independent sources before it is asserted.
Turns a vague idea into a buildable spec with acceptance criteria and named cuts.
Plans work within a budget, estimates cost before fanning out and asks before overspending.
Reads the agent station's own health from its tools and reports each finding with one fix.
Turns a script into a narrated video edit with voiceover, stills and a DaVinci Resolve timeline.
Builds a milestone study plan from the learner's real level to a concrete goal.
Answers customer questions in the company's voice and turns repeats into help docs.
Tutors you across sessions in short lessons with practice, remembering what you have learned.
Maps how a project could be attacked from its actual code and what to fix first.
Translates a document for meaning and tone, localizes idiom and keeps terminology consistent.
Packages an idea into a ready-to-shoot creator video brief with hooks, shot list and caption.
Runs the check that proves work is done and reports the result before claiming success.
Studies what you actually wrote, names your voice, then drafts in it.
Clicks through a web app like a real user, tries odd inputs and returns a ranked bug report with screenshots.
Turns a repeated read-only website lookup into a saved skill that fetches the answer directly.
Reviews last week's calendar and commitments, clears piled-up notes, flags stalled work, and plans next week around your available time.
Splits a job too big for one agent into parallel pieces, assigns each to the right agent, and merges the results into one answer.