AI agent skills portfolio: 25 Practical Ways to Automate Your Work

What you can get done with the AI agent skills portfolio

You are not buying files. The AI agent skills portfolio is built to cut the time you spend rebuilding the same workflows, and tasks that are easier to hand to your agent and to check afterwards.

• Research: check claims against sources, keep a research ledger and verify quotes.
• Development: work through package upgrades, make sense of failed builds and audit your agent's instruction files.
• Website and content: build article pages, run SEO and schema preflights, and review crawlability and AI-answer visibility.
• Video and ecommerce: turn briefs into short-video packs, launch scripts, captions and subtitles.
• Writing and learning: match your own writing voice, learn a subject step by step and think a decision through.
• Risk review: review terms and conditions, spreadsheet models and policies, and check a message for scam warning signs.

What are AI agent skills, and why use an AI agent skills portfolio?

AI agent skills are folders of plain-language instructions that tell an AI agent how to do one job well. Each skill has a SKILL.md file for the agent to read, and a Python script for the checking process. For three skills (Council of Minds, Future Self Council and Teach Me Like) that script is optional.

Without skills, you explain the job again in every chat. The agent may skip a step or forget a rule. With skills, the steps are written down once, and the agent follows them the same way each time.

This AI agent skills portfolio keeps the work split cleanly. The agent reasons about your material. The script checks the output. Most scripts use only the Python standard library, and 23 of the 25 skills work fully offline. When the script and the agent disagree, you see it.

The result is an AI agent skills portfolio of small, reliable tools. You can open every file, read every rule and see exactly what each skill reads and writes.

Inside the AI agent skills portfolio: all 25 skills

The AI agent skills portfolio falls into six groups. Pick the ones that match your work, or use all of them.

Developer & Agent Tools

• 

  DriftGuard

  AI Dependency Migration Agent

  After a package or SDK upgrade, finds the files that use it, asks a Claude model for the breaking changes and a patched version of each file, runs your tests in a temporary copy, retries and rolls back if they still fail.

• 

  PipeFix AI

  CI/CD Failure Analyzer

  Takes a failed CI/CD log and a repository folder, classifies the failure, lists the files to inspect and the kind of repair to try. Your agent writes the repair plan; it can optionally run your own tests from a fixed allow-list. It never edits, commits or deploys.

• 

  ContextGuard

  AGENTS.md & CLAUDE.md Auditor

  Audits the memory and instruction files your agent reads (AGENTS.md, CLAUDE.md and similar) for pasted secrets, injected instructions, invisible characters, risky commands, contradictions, stale entries and outside-address instructions.

• 

  SkillCheck

  Skill Folder Preflight

  Checks a skill folder or zip before you publish or sell it: frontmatter, description, missing files, placeholder text, secrets, oversized files, missing tests and examples, script problems and inconsistent versions or wording.

Learning, Thinking & Writing

• 

  Teach Me Like

  Learn Anything the Greats' Way

  Teaches a topic through the documented teaching method of a great mind (Feynman, Socrates, Einstein, Polya and more), as a quick explanation, an interactive lesson or a study pack with a reality check and a source note.

• 

  Future Self Council

  Ask the Versions of You

  Builds a panel from your own selves across time (70, in 5 years, the worried you, the bold you), lets them clash, then gives a regret check, one cheap reversible experiment, a short letter and a closing question. Claims are labelled.

• 

  Council of Minds

  Ask Any Great Mind

  Convenes the thinkers, writers or poets you name, has each apply their documented way of thinking to your dilemma, lets them clash and closes with agreement, difference, one step for this week and one question. Claims are labelled.

• 

  Voice Match

  Write Like You, Measured & Proven

  Measures your own writing style, tells you whether a draft fits it, lists exactly what differs, and checks that a rewrite kept every number, link, quote, name and yes/no. For your own voice only.

Research & Verification

• 

  ClaimLock

  Audits a draft against the sources you supply: checks each evidence passage word for word, and that numbers, dates, quotes, names, strong words, causal wording and direction match. Lists checkable sentences that have no claim.

• 

  ScoreTrace Ledger

  Tamper-Evident Track Records

  Keeps an append-only, hash-chained ledger of public forecasts and outcomes so later edits, deletions and reordering show. Reports hit rate, interval, verdict and Brier score per forecaster, and weights only calls recorded before they resolved.

• 

  ScoreTrace AI

  Forecast Track Record Auditor

  Audits the track record of a public forecaster from dated, checkable calls: hit rate with a 95% interval, a test against a baseline, calibration, returns and flags for small samples, missing losers and duplicates. No verdict under 30 resolved calls.

• 

  Source Desk

  Research Ledger & Quote Checker

  Turns a research question into a checked source ledger: verifies every quote against the saved text, counts copies as one origin, flags conflicts and old sources, and seals the ledger with a hash. Exports a folder ClaimLock can use.

• 

  OpenSignal

  Zero-Key Trend Intelligence

  Collects public, keyless signals on a topic, niche or URL from Google News, Hacker News, GitHub and Wikipedia, marks which are fresh and which sources failed. Your agent scores opportunities with a rubric and writes a brief tied to cited evidence.

Rules, Risk & Safety Review

• 

  T&C Liability Shield

  Helps you understand a Terms of Service, Privacy Policy or EULA: a script finds common clause types (arbitration, auto-renewal, data sale, liability caps and favourable ones) and quotes them with line numbers; your agent explains them plainly.

• 

  ModelTrace AI

  Spreadsheet Model Integrity Auditor

  Audits an Excel model (.xlsx or .xlsm): broken references, formulas that break their neighbours' pattern, typed numbers among formulas, short SUM ranges, hidden sheets, external links and manual calculation, with exact cells and a dependency map.

• 

  RuleBook Compiler

  Policy to Decision Rules

  Turns a written policy and your past decisions into a tested decision table: finds conflicting, unreachable, overridden and missing rules, replays past decisions, and decides new cases the same way every time.

• 

  ScamShield

  Free Scam Warning-Sign Checker

  Helps you screen a suspicious message, link, job offer, deal or payment request before acting: a local script extracts objective evidence (English, French and Arabic patterns, hidden characters, look-alike domains) and your agent explains it. It never calls anything safe.

Video Production

• 

  ReelPack Studio

  Short Video Brief to Pack

  Takes one short video from brief to a pack: ranked hooks, a timed shot list, per-shot prompts, captions, cut list, metadata and a pre-flight QA that checks claims, banned terms, likeness consent, licences and AI labelling.

• 

  LaunchLoop

  Ecommerce Launch Kit & Video Scripts

  Turns one ecommerce product into a fact-checked launch kit (truth ledger, positioning, product page, one video script, channel versions, launch sequence, test card) and checks it for wrong prices, unusable claims, false urgency and unsupported testimonials.

• 

  Short Video Caption & Subtitle QA

  Checks .srt and .vtt caption files for reading speed, line length, overlaps, timing and broken characters, and with your script, for missing, extra or changed words and numbers. Applies only mechanical fixes to a separate copy and re-checks.

Website, SEO & Marketing

• 

  Article HTML Builder

  Writes a new article from a title and keyword, or formats your draft, as clean HTML for a WordPress Custom HTML block, then runs 16 publishing checks (unsafe code, hidden text, unverified links, altered buttons, headings, contrast, placeholders).

• 

  Pagewright

  AI Website Builder

  Turns a business brief into a multi-page static website (responsive, accessible, no external requests) and audits it: links, contrast, placeholders, absolute claims, unconsented testimonials, risky forms and hand-edited files.

• 

  PagePreflight

  Free HTML & SEO Preflight Check

  Checks a local HTML page or folder for SEO and structure problems with exact line numbers (57 checks): head tags, headings, links, images, structured data, duplicated blocks and your own house rules. Your agent fixes what is safe and re-runs.

• 

  CrawlTrace AI

  AI Crawlability & Schema Auditor

  Measures whether 16 AI and search crawlers can reach your pages, then checks markup, JSON-LD and discovery files with 26 checks. Gives evidence, a per-crawler table and ready-to-paste robots.txt and JSON-LD patches.

• 

  GeoTrace AI

  Generative Search Audit

  Your agent runs queries against the AI-answer providers you choose and records each answer. A script checks the records and measures how often your brand is mentioned, recommended and cited, how competitors compare, which sources recur and where the gaps are.

Get all 25 for $39

How the AI agent skills portfolio works in three steps

1. Download. Unzip a skill into the folder your agent reads skills from. Nothing to install on a server.
2. Process. Ask your agent to do the job. It reads the SKILL.md and follows the steps.
3. Check. The skill's script checks the result and reports what passed, what failed and why. You see the check before you rely on the output.

The Quick-Start Guide walks through your first run, and the Workflow Recipes show how to chain several skills together, for example checking claims, building the article page and then auditing it.

From chaos to clarity with the AI agent skills portfolio

Most people use their agent with a fresh, vague prompt each time. The results vary, and nobody can say why one run was good and the next was not.

• Repeatable steps instead of one-off prompts.
• A script that checks the output, not just the agent's word.
• Plain files you can read, edit and keep.
• One folder per skill, so nothing is hidden.

2,196 Automated Tests Across 25 AI Agent Skills

Every skill includes its own automated test suite. You can run the tests yourself to check whether its validation scripts behave as designed. The tests provide repeatable technical checks, not a guarantee that every AI-generated result will be correct.

The tests were written by hand from known inputs and expected results. The largest suites are SkillCheck (236), CrawlTrace AI (200) and Source Desk (150). Each skill's README states its own count and the command to run it.

What the tests do and do not show

• They show the checking scripts behave as designed on the cases the tests cover.
• They let you re-run the same checks on your own machine.

They do not mean a skill will give a perfect answer on your data, and they do not replace your own judgment. OpenSignal's 28 tests run without a network connection. Most suites need only Python. DriftGuard and PipeFix AI use pytest and the packages in their requirements files.

Test count for every skill

See the AI agent skills portfolio run: three real screenshots

The text in these three images is the real output of real runs of skills in this AI agent skills portfolio, shown in a terminal-style window. The first two are trimmed to their opening lines.

Runs in Your Own Environment: Offline by Default

The AI agent skills portfolio runs in your own environment, on your own computer. There is no account to create and no hosted service to log into. Your agent and model provider have their own privacy terms, so read them.

• 23 of the 25 skills work fully offline. DriftGuard and OpenSignal use the internet.
• The documents say exactly where the internet is used, and why.
• The Data-Handling Summary lists what every skill reads, writes and sends.
• No model is included. You use and pay for your own.

Six bonus resources that come with the AI agent skills portfolio

All six are covered by the same personal-use license, which is included as a PDF.

• Quick-Start Guide for your first run.
• Data-Handling Summary covering all 25 skills.
• Workflow Recipes with real commands to chain skills.
• Skill Finder, a spreadsheet to search the 25 skills by job.
• Content Publishing Tracker for article and page work.
• Build-Your-Own-Skill Starter Kit to build a skill of your own.

Price: the AI agent skills portfolio for $39

If you bought the skills one by one at their individual prices, the total would be $248. The launch price of the AI agent skills portfolio is $39, so you save $209 (84%). The price changes on November 20, 2026. The launch price works out to about $1.56 per skill, less than the lowest individual price of $7.

The $248 figure is the total of the individual prices set for this offer. It is a comparison for this AI agent skills portfolio, not a claim about past sales.

Individual price of every skill

Get the full AI agent skills portfolio

$248

$39

Launch price. The price changes on November 20, 2026.

• 25 tested skills.
• 6 bonus resources.
• One-time payment, personal use.

Digital download. Refund terms are in the license and below.

Personal-use license and refund terms

Your purchase gives one adult a non-exclusive, non-transferable license to use the AI agent skills portfolio. Please read the full license PDF before buying.

Need client, agency or team use? Write to admin@genialthings.com about a separate commercial license.

Refunds

Because the files are digital and delivered at once, refunds are not given for a change of mind. If a file is defective or materially different from this description, report it within 7 days. I will try to fix it first, and if I cannot, you get a full refund. A duplicate purchase is refunded if reported within 7 days. Send the request to admin@genialthings.com, or use "Get Support" on your WarriorPlus purchase page. Refunds are issued through WarriorPlus, and its rules and your legal consumer rights still apply.

Requirements before you buy the AI agent skills portfolio

Please read this before purchasing. The skills are meant for an agent and model that you choose, and your setup must support the skill you want to use. Depending on the skill, that can include:

• Reading and writing files, and shell or terminal access.
• Python 3. You can get it from python.org.
• A compatible AI model or provider account.
• Internet access, for two skills only.

Special requirements

• DriftGuard: Python 3.12 or newer, internet access and your own Anthropic API key.
• OpenSignal: internet access.
• PipeFix AI: Python 3.11 or newer and the packages listed in its requirements file.
• Council of Minds, Future Self Council and Teach Me Like: work through your agent's instructions. Their Python checker is optional.

The skills are written for a Claude-style agent, and other agents were not tested. No AI model or provider subscription is included, and provider fees are separate. Compatibility with every agent, model or operating system is not promised, and you remain responsible for reviewing outputs before you rely on or publish them.

Questions about the AI agent skills portfolio