Fix admin bootstrap security and search injection vulnerabilities

[criptogus]

What does this PR add?

Fixes critical security vulnerabilities in admin account management and search functionality, plus removes accidentally committed environment files.

Type

• New package (skill / playbook / soul / guardrail)
• Improvement to an existing package (bumped version)
• Platform code or docs

Changes

Security Fixes

1. Admin Bootstrap Privilege Escalation (claimAdminBootstrap)

   • Switched from user-scoped Supabase client to service-role client for admin existence check
   • User-scoped client is blinded by RLS policies (roles read self), allowing any authenticated user to claim admin if none exists
   • Service-role client provides true global invariant check, preventing multiple admins from being created

2. Search Filter Injection (listAccounts)

   • Added sanitization of user-provided search terms before interpolation into PostgREST .or() filter
   • Strips metacharacters (%, ,, (, ), *) that could be exploited for filter injection attacks
   • Only applies filter if sanitized term is non-empty

3. Removed Committed Secrets

   • Deleted .env, .env.development, and .env.production files containing Supabase and Stripe API keys
   • Updated .gitignore to prevent future commits of environment files
   • Kept .env.example for reference

Notes for reviewers

• The admin bootstrap fix addresses a critical authorization bypass where RLS policies inadvertently allowed privilege escalation
• Search sanitization is conservative but necessary; the regex removes all PostgREST filter syntax characters
• Environment files were accidentally committed with real credentials and should be rotated immediately in production

https://claude.ai/code/session_01TSEwzP9iRBZiLkmCV2RGWx