Skills / Claude Code skills

Claude Code skills, ready made

Every kit includes your agent for Claude Code (a CLAUDE.md and a subagent) and its working skills as SKILL.md folders. Copy them in and Claude uses them when the job comes up.

Build your agent Browse ready-made agents

Claude Code skills are folders with a SKILL.md file inside. The file starts with a short header (a name and a description that says when to use it), followed by plain-text instructions. Claude Code only reads the descriptions up front and loads the full skill when your request matches, so you can keep plenty of skills without slowing anything down. You can also run one directly by typing / and its name.

Where Claude Code skills go

~/.claude/skills/<skill-name>/SKILL.md: your personal skills, available in every project.
.claude/skills/<skill-name>/SKILL.md inside a project: shared with everyone who works on that project (commit it to git).
A skill folder can also hold scripts, templates and reference files that the instructions point to.

How to create a Claude skill

Make a folder named after the job, for example .claude/skills/invoice-reminder/.
Add a SKILL.md with a header and the steps:

---
name: invoice-reminder
description: Use to chase an unpaid invoice. Drafts a polite
  reminder that gets firmer each time. Never sends without asking.
---

# Invoice reminder
1. Find the invoice number, amount and due date.
2. Check how many reminders have gone out already.
3. Draft the next reminder in the user's tone...

Restart Claude Code in that folder, then ask: "Which skills can you see?"

Tips: write the description as "Use when…" so Claude knows exactly when to reach for it; keep one job per skill; put the rules that must never break (like "never send without asking") in the skill itself.

Claude skills examples from our kits

These are real skills from KitMyAgent agents. Each comes as a ready-made SKILL.md folder:

reminder-ladder (Chase, the invoice chaser): drafts reminders that firm up step by step, from friendly nudge to final notice.
aged-debt-list: answers "who owes me money?" with an aged list of unpaid invoices, oldest first.
overnight-triage (Hollis, the night-shift assistant): sorts the overnight inbox into urgent, needs-your-OK, done and FYI, and drafts the replies.
morning-handover: one short note each morning covering what's done and what needs you.
quote-from-notes (Rivet, the trades office): turns site-visit notes or a voice note into an itemised quote from your own price list.
company-fact-sheet (Sterling, the investor's desk): a one-page fact sheet on any company, with sources and dates. Research only, no buy or sell view.
cve-triage (Vigil, the threat-intel agent): works out whether a vulnerability in the news affects you and what to do.
morning-briefing (Ashworth, the butler): a short, understated briefing of the day.

The best Claude skills to start with

The skills that pay off fastest are the boring, repeated jobs: chasing invoices, a morning briefing, triaging an inbox, turning notes into quotes and tracking follow-ups. Build them once and they're done the same way every time. Every advanced KitMyAgent agent comes with six or seven of these, and the same SKILL.md folders also work in Claude.ai, Codex, OpenClaw and Hermes.

Install in three steps

Copy claude-code/CLAUDE.md into the root of your project (or add its sections to your existing one).
Copy the claude-code/.claude/ folder into the project root. Your agent and its skills land in .claude/agents/ and .claude/skills/. For every project, use ~/.claude/ instead.
Restart Claude Code and ask: "Which agents and skills can you see?"

Menu-by-menu walkthrough: the Claude Code install guide.

Advanced agents: a full set of skills

These agents come with hand-written skills for their whole job, on top of the agent itself. 3 downloads each.

Chase · AI Invoice Chaser · 6 skills

Aged debt list: Use to answer "who owes me money?". Builds an aged list of unpaid invoices (current, 1–30, 31–60, 61–90, 90+ days overdue) with totals, the oldest first, from the user's invoices or accounting export.
Cash forecast: Use to show the user what money is expected in over the next 4–8 weeks, from unpaid invoices and agreed payment plans, with a realistic version that allows for the usual late payers.
Late payment interest: Use when a UK business customer pays late and the user wants to know what they can claim. Explains statutory late-payment interest and fixed compensation, and works out the figures. General information only.
Payment plan: Use when a customer says they can't pay in full. Drafts a fair payment-plan proposal (instalments, dates, what happens if one is missed) for the user to approve, and then tracks it.
Reminder ladder: Use to chase an unpaid invoice. Drafts reminders that firm up step by step (friendly nudge, reminder, firm, final) so the user never has to write the awkward one. Never sends without approval.

Rivet · AI Trades Office Assistant · 6 skills

Appointments and reviews: Use to draft appointment reminders the day before a visit or job and, after a job is complete and paid, a polite request for a review. Also drafts replies to reviews.
Deposits and payments: Use to track deposits and stage payments on jobs and to draft polite payment requests and reminders. Knows which jobs are waiting on money before work can start.
Enquiry triage: Use when a new job enquiry arrives (email, message, website form). Works out what's being asked, what's missing to quote, and drafts a friendly reply asking for exactly that. Never promises dates or prices.
Job tracker: Use to keep one up-to-date list of every job from enquiry to paid, and to answer questions like "what's on next week?" or "which jobs are waiting on materials?".
Quote from notes: Use to turn site-visit notes, photos or a voice note into an itemised quote using the user's own price list. Flags anything not on the list instead of guessing.

Ashworth · AI Butler · 6 skills

Errands and admin: Use for household and personal admin: booking appointments, comparing options, filling in forms, chasing refunds and organising paperwork. Prepares everything and asks before any booking, payment or message goes out.
Evening turndown: Use in the evening (or on "Anything for tomorrow?"). Prepares the user for tomorrow: an early-start warning, what to have ready, loose ends from today, and a calm sign-off.
Morning briefing: Use each morning (or on "Good morning"). Delivers a short, understated briefing of the day: diary, weather, anything urgent, and one thing worth remembering, in the butler's voice.
Occasions keeper: Use to remember birthdays, anniversaries and occasions, suggest thoughtful gifts within a budget, and remind the user in good time to buy, post or book.
Travel butler: Use when the user is travelling. Builds the itinerary, a packing list, check-in reminders and a one-page trip sheet with key times, addresses and confirmation numbers.

Hollis · AI Night-shift Assistant · 8 skills

Approvals queue: Use whenever an action needs the user's permission. Keeps one numbered queue of pending approvals, lets the user approve in bulk ("yes to 1, 3 and 5"), carries out only what was approved and reports back.
Follow up tracker: Use to keep track of promises and loose ends: replies the user is waiting for, things they promised others, and deadlines. Reminds the user before things slip and drafts polite chasers for approval.
Interrupt rules: Use whenever something new arrives while the user is busy or off duty, to decide whether to message them now, save it for the morning brief, or say nothing. Follows the user's own three-bin rules and quiet hours, and flags anything it isn't sure about.
Morning handover: Use first thing each morning (or when the user asks "what happened overnight?"). Produces one short handover note covering what's done, what needs their OK, what was held back and today's priorities.
Overnight triage: Use for the overnight or out-of-hours sweep of email and messages. Sorts everything into urgent, needs-your-OK, done and FYI, drafts the replies, and holds anything risky for the morning. Never sends without approval.
Running costs review: Use on the first Monday of each month (or on "check my agent costs") to review every background and scheduled job the user runs, what it cost and whether it earned its keep, and recommend what to keep, slow down, merge or switch off.
Weekly wrap: Use at the end of the week. Summarises what got done, what is still open, money in and out if tracked, and the top priorities for next week, in one page.

Sterling · AI Investor's Desk · 7 skills

Company fact sheet: Use when the user names a company or ticker and wants to understand it. Researches the business and its numbers and fills in a one-page fact sheet with sources and dates. Research only, no buy or sell view.
Earnings day brief: Use before or after a company's results. Before, it sets out what is expected and what would surprise. After, it compares the results with expectations and the company's own guidance, in plain English. Research only.
Eight schools check: Use when the user wants to look at a company through famous investing approaches. Runs eight checklists (principles associated with Buffett, Graham, Lynch, Greenblatt, Terry Smith, Burry, Eisman and Piotroski) and shows an agreement map. Education only, no score and no verdict.
Investment journal: Use when the user buys, sells or is thinking about a share, or sets their own investing rules. Keeps a journal of their reasons and their own trim/exit rules, and reminds them of both when prices move. Never makes decisions for them.
Red flag hunter: Use when the user shares results, an annual report, a news story or a ticker and wants warning signs checked. Looks for debt creep, cash that doesn't match profits, dilution, accounting gaps and hype. Lists findings with sources; never gives a verdict.
Weekly watchlist review: Use once a week (or when asked) to review the user's watchlist and holdings. Summarises price moves, news, results dates and any of the user's own rules that were triggered. Facts only, under one page.

Vigil · AI Threat Intel Briefer · 7 skills

Cve triage: Use when the user names a vulnerability (a CVE number, a product flaw in the news, or a vendor advisory) and asks whether they need to act. Works out whether it is exploited, whether the user is affected, what the fix is and how urgent it is.
Daily threat briefing: Use when the user asks for their threat briefing, security news or "anything I need to worry about". Checks official advisories and reputable security news, keeps only what affects the user's stack profile, and writes a short ranked briefing with sources.
Hacker trend watch: Use when the user asks what hackers are up to, for a weekly security round-up, or about threats to their sector. Summarises the week's main attack trends (ransomware, phishing, account takeovers, supply-chain attacks) in plain English, focused on businesses like the user's.
Incident first hour: Use when the user thinks they may have been hacked or scammed (a strange login, ransom note, encrypted files, money sent to the wrong account, a staff member who clicked a bad link). Calmly walks them through the first hour of defensive steps and gets the right humans involved.
Scam alert: Use when the user wants to warn staff about a current scam, or forwards a suspicious email, text or call and asks if it's genuine. Writes short staff alerts about live scams, and checks suspicious messages safely as text, without clicking anything.
Stack profile: Use first, and whenever the user mentions new software, devices or services. Builds and keeps a short profile of the technology the user's business actually runs, so every briefing can be filtered to what affects them.

Every other agent ships with its job skill too. See all 43 agents →

Questions

What is the difference between a Claude Code skill and CLAUDE.md?

CLAUDE.md is read at the start of every session, so it holds who the agent is and the rules it always follows. Skills hold how to do particular jobs and only load when needed. Keep CLAUDE.md short and put the detail in skills.

Are Claude Code skills the same as Claude.ai skills?

They use the same SKILL.md format. In Claude.ai you upload each skill as a zip; in Claude Code you copy the folder into .claude/skills/. Every kit includes both.

Do skills work as subagents?

Skills and subagents are different. A subagent is a separate helper with its own instructions (.claude/agents/<name>.md); a skill is know-how any agent can use. Every kit includes your agent as a subagent and its jobs as skills.

Can I share skills with my team?

Yes. Put them in the project's .claude/skills/ folder and commit it to git. Everyone who opens the project gets them.

Skills for other platforms: ChatGPT Claude Gemini Hermes