🛡️ Supabase Sentinel

A Claude Skill that audits your Supabase project for security vulnerabilities.

Drop it into Claude Code, Cursor, or any Claude-powered environment. Say "audit my Supabase project" and get a comprehensive security report with exact fix SQL — in minutes, not days.

170+ Lovable apps were breached. 20.1M rows were exposed across YC startups. 45% of AI-generated code introduces OWASP Top 10 vulnerabilities. Supabase's built-in Security Advisor only checks whether RLS exists — Supabase Sentinel tests whether it actually works.

---

What it does

Supabase Sentinel performs a 7-step security audit on any Supabase project:

1. Scans your codebase for exposed service_role keys, hardcoded JWTs, and secrets committed to git
2. Introspects your database schema — tables, RLS policies, views, functions, storage buckets
3. Matches against 27 known vulnerability patterns sourced from CVE-2025-48757, 10 published security studies, and thousands of documented breaches
4. Dynamically probes your API using the Prefer: tx=rollback technique (zero data modified, safe for production)
5. Tests ghost auth — can attackers create unconfirmed accounts and access your data?
6. Generates a scored security report with plain-English explanations and concrete attacker scenarios
7. Produces exact fix SQL — copy, paste, done

---

Quick start

Option 1: Claude Code / Cursor

Copy the skill folder into your project:

# Clone into your project's skills directory
git clone https://github.com/Farenhytee/supabase-sentinel.git .claude/skills/supabase-sentinel

# Or if you have a central skills directory
git clone https://github.com/Farenhytee/supabase-sentinel.git ~/claude-skills/supabase-sentinel

Then just ask Claude:

Audit my Supabase project for security issues

Claude will auto-detect your Supabase credentials from .env files, run the full audit, and present a report.

Option 2: Claude.ai (with computer use)

1. Download this repo as a ZIP
2. Upload it to a Claude.ai conversation with computer use enabled
3. Ask: "Use the Supabase Sentinel skill to audit my Supabase project"
4. Provide your Supabase URL and keys when prompted

Option 3: Manual (any AI assistant)

Copy the contents of SKILL.md into your system prompt or conversation, then follow the workflow with your Supabase credentials.

---

What it catches

Critical

| Pattern | Description | |---------|-------------| | RLS_DISABLED | Tables without Row-Level Security — fully exposed to the internet | | SERVICE_ROLE_EXPOSED | service_role key in frontend code — bypasses ALL security | | POLICIES_BUT_NO_RLS | Policies written but RLS never enabled — false sense of security | | WRITE_USING_TRUE | INSERT/UPDATE/DELETE with USING(true) — anyone can modify data |

High

| Pattern | Description | |---------|-------------| | USING_TRUE_SELECT | All rows readable by anonymous users on sensitive tables | | VIEW_NO_SECURITY_INVOKER | Views bypass RLS, running as superuser | | SECURITY_DEFINER_EXPOSED | Functions in public schema bypass RLS, callable via API | | USER_METADATA_IN_POLICY | Policies reference user-modifiable metadata — privilege escalation | | UPDATE_NO_WITHCHECK | UPDATE policies without WITH CHECK — mass assignment risk | | GHOST_AUTH | Unconfirmed email signups grant authenticated sessions | | STORAGE_NO_RLS | Storage bucket missing access control policies | | JWT_SECRET_EXPOSED | JWT signing secret leaked — can forge any user's token |

Medium

| Pattern | Description | |---------|-------------| | RLS_NO_POLICIES | RLS enabled but no policies — all access silently denied (bug) | | POLICY_NO_ROLE_SCOPE | Policy applies to all roles including anonymous | | MULTIPLE_PERMISSIVE | Multiple permissive policies OR'd — most permissive wins | | RLS_PERFORMANCE | auth.uid() not cached — performance degradation, potential DoS | | PUBLIC_BUCKET | Storage bucket publicly accessible without auth | | SENSITIVE_COLUMNS | Columns named password, api_key, etc. exposed via API | | + 9 more patterns | See references/anti-patterns.md for the full list |

---

Example output

╔════════════════════════════════════════════════════════╗
║          SUPABASE SENTINEL SECURITY REPORT             ║
╠════════════════════════════════════════════════════════╣
║  Project:   https://myapp.supabase.co                  ║
║  Scanned:   2026-03-15 14:30 UTC                       ║
║  Score:     35/100 🔴                                   ║
║  Summary:   12 tables, 8 policies, 7 findings          ║
╚════════════════════════════════════════════════════════╝

🔴 CRITICAL — users: RLS Disabled

  Risk:     Anyone on the internet can read your entire users table
  Attack:   Open browser DevTools → copy anon key → curl the API → dump all emails, names, metadata
  Proof:    curl returns [{"id":"...","email":"[email protected]",...}]

  Fix:
  ALTER TABLE public.users ENABLE ROW LEVEL SECURITY;

  CREATE POLICY "users_select_own"
    ON public.users FOR SELECT TO authenticated
    USING ((SELECT auth.uid()) = id);

🟠 HIGH — profiles: SELECT policy uses USING(true)

  Risk:     All user profiles are readable by anyone, including anonymous users
  Attack:   Enumerate all profiles via the API to harvest user data
  ...

✅ PASSING: orders, payments, invoices, subscriptions (4 tables properly secured)

---

File structure

supabase-sentinel/
├── SKILL.md                           # Core skill — 7-step audit workflow (333 lines)
├── references/
│   ├── audit-queries.md               # 20 SQL queries for schema introspection
│   ├── anti-patterns.md               # 27 vulnerability patterns with severity/detection/fix
│   ├── fix-templates.md               # SQL fix templates — 7 RLS patterns, storage, auth, prevention
│   └── vibe-coding-context.md         # CVE-2025-48757, research studies, platform analysis
├── assets/
│   └── github-action-template.yml     # CI/CD workflow for continuous monitoring
├── README.md
└── LICENSE                            # MIT

How progressive disclosure works: When Claude loads this skill, it only reads the 333-line SKILL.md initially (~5000 tokens). Reference files are loaded on-demand during specific audit steps — audit-queries.md at Step 1, anti-patterns.md at Step 2, fix-templates.md at Step 5. This keeps context usage efficient.

---

Continuous monitoring (GitHub Action)

Supabase Sentinel can generate a GitHub Action that:

• Runs on every push to supabase/migrations/
• Runs weekly (Monday 6am UTC)
• Posts findings as PR comments
• Blocks merges on CRITICAL findings

Just ask: "Set up continuous security monitoring for this project."

See assets/github-action-template.yml for the template.

---

Research backing

This skill's anti-pattern database is sourced from:

• CVE-2025-48757 — 170+ Lovable apps exposed, CVSS 9.3 (Matt Palmer, May 2025)
• Escape.tech — 2,000+ vulnerabilities across 5,600 vibe-coded apps (October 2025)
• Veracode — 45% of AI-generated code introduces OWASP Top 10 vulnerabilities (July 2025)
• Carnegie Mellon SusVibes — 82.8% of functionally correct AI code was insecure (December 2025)
• SupaExplorer — 11% of indie apps expose Supabase credentials (January 2026)
• ModernPentest — 20.1M rows exposed across 107 YC startups (March 2026)
• Wiz Research — Critical auth bypass in Base44 vibe-coding platform (July 2025)
• Supabase Security Retro 2025 — Official documentation of built-in advisor capabilities and gaps
• Supabase Splinter — All 16 official security lints mapped and extended

See references/vibe-coding-context.md for the full research breakdown.

---

What Supabase Sentinel catches that Supabase Security Advisor doesn't

Supabase's built-in Security Advisor (Splinter) runs 16 lints. Supabase Sentinel extends this with:

| Gap | What Splinter misses | Supabase Sentinel covers | |-----|---------------------|-------------------| | Policy correctness | Only checks if policies exist | Tests if they actually prevent unauthorized access | | Dynamic probing | Static analysis only | Live API testing with tx=rollback | | Ghost auth | Not checked | Tests email confirmation bypass | | Mass assignment | Not checked | Detects UPDATE without WITH CHECK + sensitive columns | | Storage config | Not checked | Audits bucket visibility and storage.objects RLS | | Codebase scanning | Not applicable | Finds service_role keys in frontend code | | Key exposure | Not checked | Detects hardcoded JWTs and committed .env files | | Column-level security | Not checked | Flags sensitive columns accessible via API | | CI/CD integration | Not available | GitHub Action for continuous monitoring |

---

Contributing

Contributions are welcome! The most valuable contributions are:

1. New anti-patterns — Found a Supabase security issue not in our database? Add it to references/anti-patterns.md with severity, detection query, fix SQL, and real-world evidence.
2. Fix template improvements — Better RLS policy patterns, edge cases, or performance optimizations in references/fix-templates.md.
3. Testing on real projects — Run Supabase Sentinel on your own Supabase projects and report false positives/negatives.
4. Platform-specific patterns — Document security patterns specific to Lovable, Bolt, Replit, or other vibe-coding platforms.

How to contribute

1. Fork this repo
2. Create a branch (git checkout -b add-new-pattern)
3. Add your changes with clear documentation
4. Submit a PR with a description of the pattern and evidence

---

Roadmap

• [ ] CLI tool — npx supabase-sentinel audit for non-Claude environments
• [ ] MCP server — programmatic access for CI/CD and dashboards
• [ ] Firebase support — extend to Firebase Security Rules auditing
• [ ] Premium dashboard — historical trending, multi-project views, Slack alerts
• [ ] VS Code extension — inline security warnings in the editor

---

Safety

Supabase Sentinel is designed to be safe for production use:

• Dynamic tests use Prefer: tx=rollback — PostgREST processes the request, evaluates RLS, returns the result, then rolls back. Zero data modified.
• Auth probes use .invalid TLD — test emails use RFC 2606 reserved domains that cannot receive mail.
• Read-only introspection — schema queries only read pg_tables, pg_policies, and information_schema. No DDL or DML.
• Open source — audit the auditor. Every query and test is visible in the source.

---

License

MIT — use it however you want, commercially or otherwise.

---