What SkillPill actually builds for you.
SkillPill converts a document you already own — a book, brand guide, methodology, SOP, or any structured text — into a SKILL.md knowledge pack. That pack is what your AI loads when you ask it a question in your domain.
This is not a chatbot, a plugin, or a GPT wrapper. It is a structured set of files that any AI can read as its briefing. The pack is yours to keep, download, share, and edit. It has no SkillPill dependency once it is built.
Here is exactly what is inside every skill pack we ship:
The full pack is delivered in three platform-specific versions: /chatgpt, /claude, and /gemini. Same content, formatted for each AI's install method. You install once per AI you use — the skill is identical in all three.
Most "add your docs" tools upload everything and hope the AI notices. A skill is engineered — every chapter has a stated purpose, and the AI's loading instructions tell it exactly where to look. This is why a skill gives specific, citation-grade answers rather than generic summaries.
SKILL.md is an open format. Anthropic, OpenAI, and Google have all adopted it. Any AI that reads SKILL.md can use your skill. You are not locked in to SkillPill or to any single AI vendor — the files work wherever the standard is supported.
Create your account — the $10 trial.
You start with the trial. It is a real account, not a demo. You build one real skill from your actual document and it ships to your inbox just like it would on any paid plan.
Here is exactly what the trial includes:
| What | Amount / detail |
|---|---|
| Duration | 7 days from sign-up |
| Credits included | 100 credits — enough for one average book (≈ 300 pages, English, Standard quality) |
| Source size cap | One file, small size. Larger documents are available on paid plans. |
| Quality available | Standard (Sonnet 4.6). Pro / Opus quality requires Plus or Premium. |
| What you keep | The skill files are yours. They do not expire if you cancel. |
| What happens on day 8 | Trial auto-converts to Starter at $29/month. You get a reminder email on day 6. |
| How to cancel | One screen, one click from Billing. No call, no form, no chat. Cancel before day 8 and no further charge is made. |
| Trial charge | $10, charged at sign-up. Refundable within 14 days — email us, no reason needed. |
We send a reminder email on day 6 — one day before the trial ends — so the conversion is never a surprise. The email tells you exactly what will be charged, how to cancel, and what you keep.
To create your account, click any "Start — just $10" button on the site. You will be taken to the sign-up screen powered by Clerk. Use an email address you check regularly — your skill delivery will arrive there in around 24 hours.
Step 1 — Upload your document.
After signing up you land on the upload screen. This is the only piece of input the build needs from you. The screen has three parts: a file dropzone, a skill-name field (you can leave this at the default), and an ownership checkbox.
Formats we accept: PDF, EPUB, DOCX, MOBI, MD, HTML, RTF. We accept what humans actually have — not a stripped-down list designed to make our pipeline easier.
The 500-page cap on Trial: the trial is capped at one small file. If you want to convert a longer book or multiple documents, Starter (300 credits / month) and Plus (1,000 credits / month) remove or raise this cap.
The ownership checkbox: you will see a checkbox that reads "I own this material and have the right to convert it." This is not bureaucratic CYA — it is the honest version of what SkillPill requires. We are building a knowledge pack from your intellectual property. We need to know you have the right to do that. Checking it confirms you own the content or have permission to use it this way.
Your source file is used only to build your skill. It is deleted automatically within 24 hours of conversion. We never use your content to train any AI model. We never share it. We never sell it. You retain full ownership of what you upload and what we ship back to you.
Once you have selected your file and ticked the checkbox, the Next button activates. Click it to move to the Confirm screen.
Step 2 — Confirm the estimate.
Before anything is charged or built, SkillPill shows you the exact credit cost for your specific document. You approve it. Then we build. If you do not approve, nothing happens.
The estimate screen shows you four things: your skill name (editable), the quality level, the speed, and the total credit cost. Here is how each one works:
Skill name
Pre-filled from your filename, cleaned up. This becomes your slash-command — for example, a skill named "brand-guide" loads in Claude Code as /brand-guide. You can rename it here before building.
Skill quality
Two options: Standard (Sonnet 4.6 — recommended for most) and Pro / Opus (deeper analysis, available on Plus and Premium plans). Standard is what the trial uses. Pro costs 2× the credits and produces a more thoroughly reasoned skill — worth it for technical or research-heavy material.
Build speed
Two options: Standard (~24h) and Real-time (30–60 minutes). Real-time costs 4× the credits of Standard. On trial, Standard is selected by default. Real-time is available on all plans for any build if you want it faster and have the credits to cover it.
How the cost is calculated
The rule of thumb: ~100 credits per 300-page English book at Standard quality. Short documents cost as little as 15 credits. Pro quality doubles the base cost. Real-time multiplies the total by 4. Your trial includes 100 credits — enough for one average Standard build, leaving a small buffer. The exact cost for your specific document is shown on this screen before anything is charged.
If the conversion pipeline fails for any reason — extraction error, QA failure, pipeline timeout — your credits come back automatically. No email needed, no form to fill out. Credits appear in your balance within minutes. If you want a cash refund instead, email us within 14 days.
Once you are happy with the name, quality, and speed, click Confirm & Build. The build starts immediately. You will be taken to the processing screen where you can track progress or just close the tab — you will get an email either way.
Step 3 — The wait.
Standard builds take approximately 24 hours. Real-time builds take 30–60 minutes. The reason Standard takes as long as it does is not a queue — it is the pipeline itself. Here is what happens during those 24 hours:
You do not need to keep the tab open. The processing page shows live stage progress if you want to watch, but the email will arrive whether you check back or not.
Emails you will receive:
- Immediately after confirming: a build confirmation email with your job details, the pipeline stage it is currently on, and a link to the install guide to read while you wait.
- ~2 hours in: an optional QA milestone email if your build reaches the Auto-QA stage — lets you know things are on track.
- When ready (~24h): the delivery email with download links for all three platform packs and a suggested first question to try.
- 48 hours after delivery (if you have not installed yet): an install nudge with the three-step path for whichever AI you use.
- Day 6 of your trial: a heads-up that the trial converts to Starter tomorrow, what it costs, and how to cancel if you do not want to continue.
Delivery day.
The delivery email has a subject line: "Your skill is ready — [Skill name]." It contains three download buttons — one for each platform pack — and a suggested first question tailored to your specific skill content.
The three packs are:
- Claude pack — used for both Claude Code (the desktop app) and Claude.ai Projects (the browser version).
- ChatGPT pack — includes knowledge files and an
instructions.txtsystem prompt, formatted for a Custom GPT or ChatGPT Project. - Gemini pack — files and system instructions formatted for a Gem or a standard Gemini conversation.
You do not need to install all three immediately. Most people start with the AI they already use daily and install the others later. The download links stay live in your account — you can come back any time.
The suggested first question: every delivery email includes one question generated from your skill's top framework and primary content. It is designed to show the difference between a generic AI answer and what your skill produces. Copy it, paste it into your AI with the skill loaded, and you will see immediately what "grounded in your work" actually means.
The delivery email also confirms that your source file has been deleted. We never trained any AI on your content, and the file is gone. If you ever want written confirmation of this, reply to the delivery email and we will send it.
After install — first questions to try.
Once the skill is loaded in your AI, the best thing you can do is ask a specific question from your actual work — not a test question, not "what do you know", but something you would genuinely want answered right now. The skill is built from your source, so the more specific and domain-grounded your question, the better the answer will be.
Here are question patterns that work well by persona, to give you a starting point:
Cite your own work
If your skill is built from your book or research: "What does [your title] say about [specific concept], and what page or section is that from?" The skill should return the exact framing from your source — your words, your logic, your page number. This is the moment it clicks: your AI can now cite your work accurately.
Voice and tone check
If your skill is built from a brand guide or style doc: "Write a one-paragraph intro for [topic] in our brand voice." Compare the output to what a generic AI would write without the skill. The difference — tone, vocabulary, the things it avoids — is the skill working.
Run your playbook
If your skill is built from a methodology, SOP, or sales playbook: "Walk me through [specific scenario] using our process." The AI should follow your decision tree, use your terminology, and reference your framework — not a generic version of the concept it learned from the internet.
If the answers seem generic or the skill is not responding with your specific content, see the Troubleshooting & FAQ guide — there are usually a few common causes and most are quick to fix.
The skill is a living document. You can re-cut it from updated source material any time (uses fresh credits), or edit the .md files yourself — they are plain text and the structure is readable. Skills are designed to grow with your work.
Where to go next.
This guide covered the end-to-end path from sign-up to first question. The other guide pages go deeper on the specific parts of the process:
- Preparing your document — format requirements, what works well and what to avoid, how to get the sharpest skill from your source without rewriting it.
- Credits, billing & top-ups — the full cost model: what things cost in credits, how the billing cycle works, what expires and what does not, and how to top up.
- Installing your skill — the step-by-step install guide for ChatGPT, Claude Code, Claude.ai Projects, and Gemini. 60 seconds per AI.
- Using your skill day-to-day — how to work with citations, how to keep the skill up to date, and what good looks like in daily use.
- Troubleshooting & FAQ — failed builds, generic answers, install problems, billing questions, and everything else that can go wrong, with specific fixes.
If something is not covered in the guide, email [email protected]. A founder reads these. We respond within one business day, usually same-day.