agentleFS
chenson42chenson42/stackkeel/.claude/skills/personalize/SKILL.md

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

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
  1. Distinctive:

    Ask ONE at a time; echo the full set back for confirmation before editing anything:

    line 35
  2. Very specific:

    docs-site: Read the registry's warnings for anything the user should hear before deciding (e.g.

    line 49
  3. Distinctive:

    Not offered (core, non-removable): auth core, the permission model, flag infrastructure, audit writes, the brand engine, the email queue …

    line 57
  4. Very specific:

    If verification fails: git checkout -- .

    line 74
  5. Distinctive:

    Ask for the brand seed color (hex).

    line 82
  6. Very specific:

    Ask which of the four type pairings (classic / modern / warm / contemporary — see BRANDING.md) and record the choice in …

    line 88
  7. Very specific:

    Leave kit.name/kit.canonicalUrl UNCHANGED — they identify the upstream kit, not this fork.

    line 160
  8. Very specific:

    Run pnpm kit:check --why (expect PERSONALIZED-OK) and pnpm check.

    line 163

“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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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 it
  • scripts/kit/identity-files.json — every file the find-replace pass must touch
  • kit.json — the manifest this skill writes at the end

Step 0 — Preconditions

  1. git status --porcelain must be empty. Dirty tree → stop; ask the user to commit or stash first (personalization must be one reviewable diff).
  2. 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.json records.
    • UNPERSONALIZED / DECLINED → proceed.
  3. Confirm origin points 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:

  1. Project name (display, e.g. "Acme Portal")
  2. Slug (kebab-case; becomes root package name, app scheme, default bundle-id stem)
  3. One-line description
  4. Repository URL (the fork's own)
  5. Contact / from-email (transactional sender)
  6. 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 module

If 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

  1. Ask for the brand seed color (hex). Default offered: keep Starter Blue.
  2. pnpm brand:generate <hex> — show the user both scheme ramps and the adjustments[] log (what the generator changed to clear contrast floors).
  3. Apply: update the static --brand-* values in packages/ui/src/theme.css to the generated light/dark ramps (the runtime branding DB row stays empty until they choose to use the admin editor; the flag ui.brand_theming stays off).
  4. Ask which of the four type pairings (classic / modern / warm / contemporary — see BRANDING.md) and record the choice in docs/product/branding.md (wiring fonts is a post-personalization task; the kit stores the choice).
  5. 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: false
  • modules: 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 if gh can fetch it (gh api repos/<canonical>/commits/main --jq .sha), else leave empty for /upstream-sync first-run bootstrap
  • Leave kit.name/kit.canonicalUrl UNCHANGED — 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.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.