builder-setup
Skill1 starsChanged 23 days agoMITView on GitHub
First-time setup for Builder Kit on this computer: checks the tools it needs (Node, Git, Chrome, Python), interviews the user briefly to create their profile (how technical they are, what they build, which accounts they have), and sets up research scraping (a free local reader, paid scraper keys, or both) plus an optional signed-in research browser for Instagram, LinkedIn and X. Use it right after Builder Kit is installed, whenever the session-start notice says Builder Kit isn't set up, when research reports no working scraper, or when the user says "set up builder kit", "setup", "add my scraper keys", "connect firecrawl/apify", or "my profile".
- Reads credentials
In this repository7 other files it gives its agents.
- appSkill
- backend-loginSkill
- buildSkill
- project-memorySkill
- researchSkill
- voice-agentSkill
- websiteSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Run node "<kit>/skills/builder-setup/scripts/check-env.mjs".
- Distinctive: line 18
Ask, in plain words, with the recommended option first:
- Very specific: line 24
Write ~/.claude/builder/profile.md from references/profile-template.md.
- Very specific: line 36
Free: run node "<kit>/skills/research/scripts/research.mjs" setup.
- Very specific: line 38
Keys: create ~/.claude/builder/scrape/.env from references/env-template.txt, then open it for the user (Windows: notepad "<path>", Mac: …
- Very specific: line 49
If yes: node "<kit>/tools/browser-login.mjs".
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Builder Kit setup
About 5–10 minutes, once per computer. It exists because research without a real scraper produces average
work, and because the kit has to talk to this person at their level. <kit> below means ~/.claude/skills/builder-kit.
Step 1: Check the computer
Run node "<kit>/skills/builder-setup/scripts/check-env.mjs". For each ❌, show the user the one-line fix it prints.
Installing software is their decision: ask once, listing everything missing, and run the install commands only after
they say yes. Re-run the check afterwards. Node 22+, Git and Chrome are required. Python is needed for research.
Step 2: Profile (one question round)
Ask, in plain words, with the recommended option first:
- What should I call you?
- How technical are you? (I'm new to this (Recommended) / I know a bit / I'm comfortable with code)
- What will you build most? (multi-select: websites · apps or dashboards · logins and backends · voice or chat agents · other)
- Which accounts do you already have? (multi-select: GitHub · Vercel · Supabase · Google Cloud · Cloudflare · a domain · ElevenLabs · none yet)
Write ~/.claude/builder/profile.md from references/profile-template.md. The session-start hook shows it to every
future session, so keep it short (under 30 lines). Fill in their computer and terminal from step 1 (so commands match it). Add their screen size if they mention it. If they're experienced and want to pick design references themselves, set "Design references: I choose them".
Step 3: Research reader (required)
Research reads real pages, transcripts and threads. Ask:
How should research read the web?
- Free reader on this computer (Recommended). Installs a free page reader and YouTube transcript tool. Takes about 5 minutes and about 500 MB. No accounts needed.
- My own paid scraper keys. Use Firecrawl or Apify accounts you already have (both have free monthly credit).
- Both. Free first; paid only when a site blocks the free reader.
- Free: run
node "<kit>/skills/research/scripts/research.mjs" setup. It finds or fetches a suitable Python on its own (via uv if needed). If it stops, show the user the single command it printed. - Keys: create
~/.claude/builder/scrape/.envfromreferences/env-template.txt, then open it for the user (Windows:notepad "<path>", Mac:open -e "<path>") and let them paste their keys and save. Never ask for keys in chat, and never print or echo them. Where to get keys: Firecrawl at firecrawl.dev (free monthly credits); Apify at apify.com (about $5 of free credit every month). - Then run
research.mjs doctor. At least one page reader must show UP. Runresearch.mjs statusif keys were added.
Step 4: Signed-in research browser (optional, recommended for design and social research)
Instagram, LinkedIn, X, Pinterest and Dribbble hide most content from logged-out visitors. Offer:
"Want research to read Instagram, LinkedIn, X and Pinterest? I'll open a separate Chrome window. Sign in there with a spare account, not your main one, because automated reading can get an account flagged. Then close the window."
If yes: node "<kit>/tools/browser-login.mjs". The free reader uses that saved window's sign-ins for those sites
from then on. It only ever reads; it never posts, likes or messages.
Step 5: Prove it works
- Screenshot test:
node "<kit>/tools/shot.mjs" --url https://example.com --size 390x844 --out "<temp>/bk-test.png", then read the PNG. - Research test:
research.mjs get https://example.com. It should print the page text. - Tell the user, in three lines: what's ready, what's optional and skipped, and how to start ("Just tell me what you want to build, like 'make me a website for my bakery'").
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

