firebase-security-rules-auditor Skill for Claude: Red-Team Your Firebase Rules

firebase-security-rules-auditor, from Google's Firebase team, makes your agent audit Firestore and Storage rules and score them 1 to 5.

Installs: 142.2K; on skills.sh
Stars: 497; on GitHub
Updated: Oct 2026; last repo commit

firebase-security-rules-auditor sets the agent up as a senior security auditor for Firebase Security Rules. It is told not to trust a rule because it looks complex and to look for a sequence of operations that gets around it.

It works through a six-point checklist: update bypass, authority source, business logic, storage abuse, type safety, and field-level versus identity-level security. It then scores the rules from 1 (critical) to 5 (secure) and returns JSON with a score, a summary and findings that each have a check, severity, issue and recommendation.

The skill is one SKILL.md with no commands and no scripts.

Good for

Developers who want a security review of firestore.rules or storage.rules
Reviewers who need a repeatable checklist and a score
Teams checking that rules fit the app's business logic

Install firebase-security-rules-auditor

Claude Code

Copy this prompt.

Install the firebase-security-rules-auditor skill in Claude Code. Run this command:

claude plugin marketplace add firebase/agent-skills
claude plugin install firebase@firebase

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js 20+, Google account.

The plugin route is the one the Firebase skills recommend for Claude Code. It installs all 13 skills in the repo and a Firebase MCP server, which .mcp.json starts with npx -y firebase-tools@15.32.1 mcp --dir . (version at review time). Restart Claude Code afterwards. The skills CLI route installs only the skills named. Every route needs Node.js 20 or newer, and the skills run the Firebase CLI through npx, so you also sign in to Firebase with a Google account. The README writes the marketplace as firebase/skills; the repository is firebase/agent-skills, the name the vendor Claude Code setup guide uses.

From the vendor's docs. Source.

What's in the skill

When it loads

Audits Firebase (Firestore, Cloud Storage) security rules for vulnerabilities, privilege escalation, role bypasses, create vs update inconsistencies, resource exhaustion, type safety, size limits, and hasOnly ownership checks. Use when auditing/reviewing rules, running red-team rule assessments, or scoring against auditor checklists. Don't use for Firebase CLI (login, deploy), Auth, Crashlytics, Remote Config, or database queries.

Sections in SKILL.md

Assessment: Security Validator (Red Team Edition)
Mandatory Audit Checklist:
Admin Bootstrapping & Privileges:
Scoring Criteria (1-5):

Key facts

Author: firebase (GitHub: firebase)
Repo: firebase/agent-skills
License: Apache-2.0 (LICENSE and README); the Claude plugin.json says MIT
Installs on skills.sh: 142,227 as of 2026-11-02
GitHub stars (whole repo): 497 as of 2026-11-02
Last repo commit: 2026-10-30
Files: 1 file (SKILL.md), no scripts
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk pass
Made by: Google LLC (SKILL.md metadata); the Claude plugin calls itself the official Claude plugin for Firebase
Category: CloudSecurity

What it includes

Can a valid create be updated into a bad state?
Compares create and update rules for changes to roles, size limits or data types.
1. The Update Bypass
Does it trust user-supplied role or owner fields?
Flags rules that rely on request.resource.data for fields like role, isAdmin or ownerId.
2. Authority Source
Do the rules support what the app does?
For example, whether collaborators in a sharing app can actually read the data.
3. Business Logic vs. Rules
String length and array size limits
Missing limits are labelled a resource exhaustion or DoS risk.
4. Storage Abuse
is string, is int, is timestamp checks
Checks that fields are type-checked.
5. Type Safety
hasOnly() and diff() need an ownership check too
These limit which fields change, not who changes them. Without a check like resource.data.uid == request.auth.uid it is a data integrity hole.
6. Field-level vs identity-level
One hardcoded admin email is allowed if verified
A single admin email does not count against the score if email_verified is also checked and no one can add themselves as admin.
Admin bootstrapping
1 Critical to 5 Secure
1 is unauthorized data access or privilege escalation, 2 is broken business logic, 3 is PII exposure or create/update inconsistency, 4 is minor issues, 5 is strict and complete. Output is JSON with score, summary and findings.
Scoring and JSON output

From the official docs

Permissions

Low risk

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

The plugin route adds a Firebase MCP server
The Firebase plugin ships an .mcp.json that starts `npx -y firebase-tools@15.32.1 mcp --dir .` with IS_FIREBASE_MCP=true, so a Firebase MCP server runs in your project folder. The pinned version was 15.32.1 at review time. The plugin also installs all 13 skills in the repo. The skills CLI route installs only this skill and adds no server. github.com
The skill itself is one instruction file
The SKILL.md is a checklist and scoring rubric. It names no commands, scripts or API keys, so on its own it only changes how your agent reviews the rules you give it. github.com

Questions

What does firebase-security-rules-auditor do?

It makes your agent review Firebase Security Rules for Firestore and Cloud Storage like a penetration tester. It follows a six-point checklist, scores the rules from 1 to 5 and returns the findings as JSON.

How do I install it in Claude Code?

Use the install box above. The plugin route installs every Firebase skill and the MCP server. The skills CLI route installs only this skill.

Does it fix my rules?

It audits and recommends. Each finding carries a recommendation. Writing rules is a different job: the firebase-firestore skill sends rule authoring to the firestore-rules-creation skill.

Does it need an API key or the Firebase CLI?

The SKILL.md names no key, no command and no CLI. It is a checklist and scoring guide for the agent.

Is it safe to use?

It is one SKILL.md with no scripts and no commands. Its job is to read your rules and report. The score is the model's judgment against a rubric, so treat it as a review aid and not a guarantee.

What does a score of 3 mean?

Moderate: PII exposure such as public emails, or inconsistent validation between create and update on critical fields. A 1 is unauthorized data access or privilege escalation, and a 5 is comprehensive validation with strict ownership.

Information

updated Nov 2, 2026

Repository: firebase/agent-skills
Folder: skills/firebase-security-rules-auditor

firebase/agent-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 6

agent-skills README · github.com
firebase-basics Claude Code setup · github.com
Firebase plugin .mcp.json · github.com
firebase-basics Cursor setup · github.com
firebase-security-rules-auditor SKILL.md · github.com
github.com · github.com