A prompt that maps your actual coding abilities through a short conversation with Claude, then generates a portable skill profile you can drop into any AI coding project.
AI coding assistants don't know what you know. Start a new project and Claude might explain what an API is when you've built five of them. Or it might skip over a concept you've never seen before because it assumed you'd follow along. This gets worse the more projects you work on, because context doesn't carry between them.
This is especially true for people learning to code with AI. Your skills change fast. What you could do last month isn't what you can do today. And "I built it with AI help" covers a huge range, from "I understand every line" to "I pasted the error and it fixed it."
Drop the prompt into a Claude conversation. It runs a 5-10 minute adaptive assessment (conversational, not a quiz) and generates a skill profile block like this:
## Developer Skill Profile
Generated: March 2026
### Summary
Bill is a self-taught developer three months into coding, building production
multi-tenant SaaS applications across Node.js/Express and Python stacks...
### Calibration Notes
When working with Bill:
- Do not assume he can read code fluently. When showing or explaining code,
say what each meaningful block does and why.
- Do explain architecture decisions. He reviews plans for reasonableness and
wants to understand the "why"...
- Do assume he understands the general shape of web applications. He doesn't
need "what is an API" explanations, he needs "here's why we're structuring
this API route this way" explanations...
You paste that block into your project's custom instructions (or Claude's memory, or a CLAUDE.md file, or whatever context mechanism your AI tool uses). Now every conversation in that project starts with the right calibration.
- Copy the contents of
skill-profile-prompt.md - Paste it into a new Claude conversation, or into a Claude Project's custom instructions
- Say "Let's do a skill assessment"
- Answer the questions honestly (5-10 minutes)
- Copy the generated profile into your projects
Re-assess periodically. If you're actively learning, your profile goes stale. Run it again after a major project or every couple of months.
The generated profile is just a block of text. It works anywhere your AI tool accepts context:
- Claude Projects: Paste into the project's custom instructions
- Claude Code: Add to your
CLAUDE.mdfile - Cursor / Windsurf: Add to your rules or context files
- Any AI chat: Paste at the start of a conversation
The assessment maps four areas, adapting its questions based on your answers:
- Coding fluency ... can you read, write, and debug code? At what level?
- Tools and workflow ... what's your dev environment? Git comfort? Terminal comfort?
- Architecture concepts ... do you understand APIs, databases, auth, deployment?
- AI-assisted development ... how do you work with AI tools? Do you use structured context?
It distinguishes between "I built this" and "AI built this while I directed it," because both are valid but they mean different things for how an AI assistant should talk to you.
This is a living prompt. If you run the assessment and find gaps, or if you adapt it for a different AI tool, PRs are welcome. A few guidelines:
- Keep the tone conversational. This isn't a certification exam.
- The prompt should work for complete beginners and experienced developers. Don't add questions that assume prior knowledge.
- The Calibration Notes section of the output is the whole point. Any changes should ultimately make that section more useful.
MIT. Use it, fork it, build on it.