personalize
Skill1 starsChanged 52 days agoMITView on GitHub
One-session rebrand of a fresh kit copy into a real project — identity interview, app/feature selection with registry-driven stripping, brand generation, strategic docs, manifest write. Run when kit-check reports UNPERSONALIZED. Never commits; the user reviews the diff.
- Deletes or force-pushes
4 other projects use a version of this.
Adopted by
Most recently taken up by
- duncan-buildroom/profit-room-skillstheir own version, first seen Oct 13, 2026
- duncan-buildroom/freeskillstheir own version, first seen Oct 13, 2026
- palashjain95/jobhuntertheir own version, first seen Mar 28, 2026
- aleksander-dytko/ai-pm-workspacetheir own version, first seen Feb 27, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- duncan-buildroom/freeskills★ 65same name
- aleksander-dytko/ai-pm-workspace★ 34same name
- duncan-buildroom/profit-room-skills★ 18same name
- palashjain95/jobhunter★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- add-permissionSkill
- downstream-syncSkill
- merge-prSkill
- neon-postgresSkill
- new-featureSkill
- pre-pushSkill
- process-feedbackSkill
- release-notesSkill
- testSkill
- trivialSkill
- upstream-syncSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 3 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
3
Distinctive
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
- Distinctive: line 35
Ask ONE at a time; echo the full set back for confirmation before editing anything:
- Very specific: line 49
docs-site: Read the registry's warnings for anything the user should hear before deciding (e.g.
- Distinctive: line 57
Not offered (core, non-removable): auth core, the permission model, flag infrastructure, audit writes, the brand engine, the email queue …
- Very specific: line 74
If verification fails: git checkout -- .
- Distinctive: line 82
Ask for the brand seed color (hex).
- Very specific: line 88
Ask which of the four type pairings (classic / modern / warm / contemporary — see BRANDING.md) and record the choice in …
- Very specific: line 160
Leave kit.name/kit.canonicalUrl UNCHANGED — they identify the upstream kit, not this fork.
- Very specific: line 163
Run pnpm kit:check --why (expect PERSONALIZED-OK) and pnpm check.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Personalize
Turn an unpersonalized copy of the kit into this fork's own project. Interview → select → strip → rebrand → document → record. Nothing is committed — the final diff is the user's to review. Every step below is a plain checklist any Agent Skills-compatible assistant (or a human) can execute.
Machine-readable inputs this skill drives from (never inline their contents here):
scripts/kit/module-registry.json— what each optional module is and how to strip itscripts/kit/identity-files.json— every file the find-replace pass must touchkit.json— the manifest this skill writes at the end
Step 0 — Preconditions
git status --porcelainmust be empty. Dirty tree → stop; ask the user to commit or stash first (personalization must be one reviewable diff).pnpm kit:check --why:CANONICAL-OK→ refuse: "This is the canonical kit repo; personalization is for copies." Stop.PERSONALIZED-OK→ confirm the user wants a re-run. Re-runs are merge-semantics: re-ask only what they want to change, never re-strip an already-stripped module, append (don't overwrite)kit.jsonrecords.UNPERSONALIZED/DECLINED→ proceed.
- Confirm
originpoints at the user's own repository (not the canonical URL).
Step 1 — Identity interview
Ask ONE at a time; echo the full set back for confirmation before editing anything:
- Project name (display, e.g. "Acme Portal")
- Slug (kebab-case; becomes root package name, app scheme, default bundle-id stem)
- One-line description
- Repository URL (the fork's own)
- Contact / from-email (transactional sender)
- License (keep MIT / proprietary / other)
Step 2 — Project selection
portal is mandatory (the kit's spine). For each optional app, present its
label + operatingCost from the module registry and ask keep/strip
(multi-select where the assistant supports it): admin, shell, mobile,
docs-site. Read the registry's warnings for anything the user should hear
before deciding (e.g. stripping admin removes the only users/roles UI).
Step 3 — Feature selection
Same, for feature modules: helpdesk, feedback, twoFactor, oidc, emailQueue, whatsNew, flagsAdmin, auditViewer.
Not offered (core, non-removable): auth core, the permission model, flag
infrastructure, audit writes, the brand engine, the email queue itself, migrations
discipline. Note that twoFactor/oidc are keep-dormant strips (registry
warnings explain) and flagsAdmin/auditViewer/emailQueue strip admin pages
only.
Step 4 — Strip
For each deselected module, in this order — apps first, then features, so shared-seam fences are removed before feature passes re-read the same files:
pnpm kit:strip -- <module> --dry-run # show the user the plan
pnpm kit:strip -- <module> # execute (also records kit.json paths.removed)
pnpm turbo typecheck build # verify after EACH moduleIf verification fails: git checkout -- . && git clean -fd <the module's paths> to
roll back, re-select the module as kept, report the failure verbatim, and continue
with the remaining modules. A SEAM PENDING note in the plan means the registry
expects 1–2 dangling imports — fix exactly what the typecheck names, nothing more.
Stripping both shell and mobile auto-strips device-auth; the dry-run shows it.
Step 5 — Branding
- Ask for the brand seed color (hex). Default offered: keep Starter Blue.
pnpm brand:generate <hex>— show the user both scheme ramps and theadjustments[]log (what the generator changed to clear contrast floors).- Apply: update the static
--brand-*values inpackages/ui/src/theme.cssto the generated light/dark ramps (the runtimebrandingDB row stays empty until they choose to use the admin editor; the flagui.brand_themingstays off). - Ask which of the four type pairings (
classic/modern/warm/contemporary— see BRANDING.md) and record the choice indocs/product/branding.md(wiring fonts is a post-personalization task; the kit stores the choice). - Light/dark/auto preference → set the default theme in both apps' providers.
Step 6 — Strategic docs
For each of docs/product/vision.md, business-plan.md, branding.md offer three
paths: (a) user pastes existing content; (b) interview — 4–6 questions each
(vision: who is it for, the one-sentence promise, what it replaces, success in 12
months; business plan: revenue model, first 10 customers, cost drivers, riskiest
assumption; branding: voice adjectives, audience register, words to avoid); (c)
skip — write the template with TODO(owner): markers.
Step 7 — Identity find-replace
Walk scripts/kit/identity-files.json (READ the registry — never a hardcoded list).
For each file, apply the note-described edits: kit name/slug → the new identity,
com.example.<kit> bundle ids → the user's reverse-domain id, deep-link scheme, docs
URLs, the docs.yml repository guard (or delete docs.yml if docs-site was stripped),
LICENSE holder. The Android package directory rename is included (registry note
on MainActivity.java). Generated Capacitor configs: edit capacitor.config.ts then
tell the user to run npx cap sync in apps/shell rather than hand-editing the
generated copies. Then sweep:
grep -ri "<kit-name>" --exclude-dir=node_modules --exclude-dir=.git . | grep -v docs/Anything left outside docs/ (history is fine to keep) gets fixed now.
Step 8 — Wire strategic docs into the instructions
Append a ## Project Context section to AGENTS.md (project name, one-liner,
pointers to docs/product/*.md) and add docs/product/vision.md to the analyst
agent's Phase-1 reading list in .claude/agents/analyst.md.
Step 9 — Manual-TODO report
Print a checklist of what only a human can do. Base set: provision Postgres (Neon) +
set DATABASE_URL; openssl rand -base64 32 → AUTH_SECRET; same →
AUTH_TOTP_ENCRYPTION_KEY (warn: cannot rotate later, if twoFactor kept); Google
OAuth client + redirect URIs per kept app; INITIAL_ADMIN_EMAILS; Resend key +
verified sender; NEXT_PUBLIC_APP_URL/AUTH_URL per deploy target; Upstash (or
accept in-memory rate limiting); Turnstile keys (or leave no-op); icon/wordmark asset
replacement beyond the generated ones; create the GitHub repo + push; after the
first push, open the repo's Actions tab and confirm the CI workflow actually ran on
your default branch — instructions in this kit claim CI enforcement, and a
workflow that exists but never triggers makes that claim silently false (the
check-instructions tripwire catches the empty-directory case, but only you can
confirm the first real run); hosting setup (e.g. Vercel projects per kept app). Conditionals: shell kept → set the deployed
portal URL for npx cap sync, Apple/Google dev accounts, real bundle ids in stores;
mobile kept → EAS account, EXPO_PUBLIC_API_URL; docs-site kept → enable
GitHub Pages on the new repo (the workflow guard was retargeted in Step 7); oidc
kept → issuer/client env vars; emailQueue kept → CRON_SECRET + cron
enablement on the host.
Step 10 — Write the manifest
Update kit.json:
identity:role: "fork",personalized: true,personalizedAt: <today>,projectName,slug,personalizationDeclined: falsemodules: the Step 2/3 selections (kept = true)paths.personalized: every file actually edited in Steps 5–8 (seed from the identity registry + theme.css + docs/product/)paths.removed: already appended by each strip (verify entries exist)sync.upstream.lastSyncedDate: today (primes the 14-day cadence clock);baselineSha/lastSyncedSha: today's canonical main SHA ifghcan fetch it (gh api repos/<canonical>/commits/main --jq .sha), else leave empty for/upstream-syncfirst-run bootstrap- Leave
kit.name/kit.canonicalUrlUNCHANGED — they identify the upstream kit, not this fork.
Run pnpm kit:check --why (expect PERSONALIZED-OK) and pnpm check.
Step 11 — Do not commit
Print a summary (identity, kept/stripped modules, files touched, TODO count) and
stop. The user reviews git diff and commits when satisfied.
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.

