Skip to main content
Agent skill

Google Cloud Filestore Auditing

Audits Filestore instances on Google Cloud across projects for disaster recovery readiness (missing or stale backups), security access governance (overly permissive NFS export rules, 0.0.0.0/0 exposure, missing ROOTSQUASH), and reliability compliance (Physical Zone Isolation PZI and Physical Zone Separation PZS). Use when assessing storage health posture, auditing NFS export permissions, identifying unprotected file shares, or validating zone failure domains. Don't use for Cloud Storage buckets, Persistent Disk, or NetApp Volumes.

Get started

Get the complete skill folder, including its scripts and reference files.

When to use this skill

Use Google Cloud Filestore Auditing when an AI agent needs a reusable SKILL.md workflow for this job: Audits Filestore instances on Google Cloud across projects for disaster recovery readiness (missing or stale backups), security access governance (overly permissive NFS export rules, 0.0.0.0/0 exposure, missing ROOTSQUASH), and reliability compliance (Physical Zone Isolation PZI and Physical Zone Separation PZS). Use when assessing storage health posture, auditing NFS export permissions, identifying unprotected file shares, or validating zone failure domains. Don't use for Cloud Storage buckets, Persistent Disk, or NetApp Volumes.

When not to use it

Skip Google Cloud Filestore Auditing when the task is outside the coding category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

Where will you use it?

Install scope
Claude Code folder setup
mkdir -p ~/.claude/skills/google-cloud-filestore-auditing
# Save SKILL.md to ~/.claude/skills/google-cloud-filestore-auditing/SKILL.md
  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/google-cloud-filestore-auditing/SKILL.md. For a shared project, commit the folder at .claude/skills/google-cloud-filestore-auditing/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

Try a starter prompt

After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.

Starter prompt
Use the Google Cloud Filestore Auditing skill for this task: Use Google Cloud Filestore Auditing when this documented workflow matches the task. Confirm any required inputs or files before making changes.

What supports this recommendation?

71/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
77/100
Practical detail
68/100
Source evidence
63/100
Source signals
72/100
Model compatibility evidence
  • claudeuntested

    No model-specific signal or recorded compatibility test was found.

  • chatgptuntested

    No model-specific signal or recorded compatibility test was found.

  • geminiinferred

    The skill text mentions Gemini or a closely associated term. This is keyword-derived compatibility; no hands-on model test is recorded.

  • copilotuntested

    No model-specific signal or recorded compatibility test was found.

  • llamauntested

    No model-specific signal or recorded compatibility test was found.

  • perplexityuntested

    No model-specific signal or recorded compatibility test was found.

  • mistraluntested

    No model-specific signal or recorded compatibility test was found.

  • grokuntested

    No model-specific signal or recorded compatibility test was found.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

What this skill does

# Filestore Auditing Skill

This skill enables autonomous agents to audit, evaluate, and report the disaster recovery, security access governance, and architectural reliability posture of Filestore fleets across Google Cloud projects.

## Prerequisites / IAM Requirements & Permissions

Before executing this skill, the runtime principal (user account or Service Account) must possess the following IAM roles and granular permissions on the target GCP project(s):

### 1. Audit Operations (Read-Only Assessment)

Requires the **`roles/file.viewer`** role, which provides:

- **`file.instances.list`**: Enumerate Filestore instances across project locations. - **`file.instances.get`**: Inspect instance configuration, NFS export rules, IP ranges, squash mode, and PZI/PZS isolation status. - **`file.backups.list`**: Enumerate existing backups across regions. - **`file.backups.get`**: Inspect backup timestamps, source instance URIs, and status.

### 2. Remediation Operations (Backup Creation)

Requires the **`roles/file.editor`** (or `roles/file.admin`) role, which provides:

- **`file.backups.create`**: Create on-demand baseline backups for unprotected instances. - **`file.operations.get`**: Monitor long-running backup creation operations.

### 3. MCP Tool Invocation

If invoking capabilities via the Filestore MCP Server (`file.googleapis.com/mcp`):

- **`roles/mcp.toolUser`**: Required to execute MCP tools (`list_instances`, `get_instance`, `list_backups`, `get_backup`, `create_backup`).

### 4. Cloud Authentication Setup

Authenticate the runtime environment using the appropriate method for your execution context:

- **CLI Execution (`gcloud`)**: ```bash gcloud auth login ``` - **Application SDKs & Scripts**: ```bash gcloud auth application-default login ``` - **Automated Services & Service Accounts**: ```bash export GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account-key.json" ```

### 5. Billing Account Attachment

- **Active Billing Required**: The target GCP project must have an active Cloud Billing account attached. Filestore instances and backup operations require active billing. Verify billing status with: ```bash gcloud beta billing projects describe {project_id} ```

## Quick Start

1. Ensure `gcloud` is installed. See [gcloud installation guide](https://docs.cloud.google.com/sdk/docs/install-sdk.md.txt) if needed. 2. Authenticate your environment:

```bash gcloud auth login gcloud auth application-default login ``` 3. Set the target project context and verify an active billing account is attached:

```bash gcloud config set project {project_id} gcloud beta billing projects describe {project_id} ``` 4. Configure a well-supported default region (e.g., `us-central1`). See official [region availability documentation](https://cloud.google.com/about/locations):

```bash gcloud config set compute/region us-central1 ``` 5. Verify required service APIs are enabled:

```bash gcloud services enable file.googleapis.com --quiet ``` 6. Run fleet discovery and backup inventory (see "Discovery & Read Operations" below). 7. Evaluate findings against the audit rules matrix and render the executive scorecard and inventory table. 8. If unprotected instances are identified, present baseline backup creation commands and prompt the user for confirmation.

## Attribution Guardrail

Tag every Cloud Filestore command or API request provided or executed. Prefix `gcloud` commands with the designated metrics environment:

```bash CLOUDSDK_METRICS_ENVIRONMENT="gcs-skills gcs-skills/1.0 (skill:google-cloud-filestore-auditing)" \ gcloud filestore instances list ... ```

On direct HTTP calls to the GCP REST API, append the `User-Agent`:

``` User-Agent: gcs-skills/1.0 (skill:google-cloud-filestore-auditing) ```

## Conceptual & Informational Queries (CRITICAL)

For purely conceptual, educational, or architectural questions (e.g., *"What is Physical Zone Isolation (PZI) in Filestore?"*, *"Why is NO_ROOT_SQUASH dangerous?"*, *"Explain Filestore backup architecture"*):

- **Rule**: Answer immediately using pre-trained knowledge and the guidance in `references/`. - **Constraint**: **Do NOT execute external tool calls or API requests** for basic conceptual queries.

## Handling "No-Command" Constraints & Evaluations (CRITICAL)

If the user prompt contains constraints like *"Do not execute commands"*, *"without executing"*, or *"read-only"*:

- **Rule**: **Strictly avoid calling the `run_command` tool** to execute any shell, python, or `gcloud` commands. - **Discovery Hierarchy**: 1. First, check if Filestore MCP tools (`list_instances`, `list_backups`) are available and query them directly (these are API invocations, not shell command executions). 2. If MCP tools are not present or cannot connect, search local reference markdown files (specifically the mock fleet definitions in `references/zone-isolation-pzi-pzs.md`) for any mock instances or project details matching the request. (Do NOT attempt to read evaluation config files such as `EVAL.yaml` or `EVAL.txtpb` during evaluation runs as access is restricted and triggers anti-cheating timeouts). 3. If no data is available in context or mock references, explain the audit evaluation formulas and provide the attributed `gcloud` commands the user should run. - **Mandatory User Confirmation Requirement**: Even when the user prompt asks not to execute commands or asks only for audit recommendations, any response recommending backup remediation MUST STILL end with a clear confirmation prompt before execution (e.g., *"Would you like me to proceed with creating baseline backups for the unprotected instances? Please confirm to proceed."*).

--------------------------------------------------------------------------------

## Core Operational Workflow

### 1. Discovery & Read Operations

The agent must discover all Filestore instances and backups in the target project.

- **Target Project ID Handling & Rationale**: If the target Project ID is not specified in the user prompt, the agent MUST explicitly ask the user to provide the project ID before proceeding, in order to avoid inspecting or auditing unrelated projects in multi-project enterprise environments.

Choose the discovery method matching your runtime environment:

#### Option A: Filestore MCP Tools (Recommended when MCP is mounted)

1. **Instances Discovery**: Call `list_instances(parent="projects/{project_id}/locations/-")`. 2. **Backups Discovery**: Call `list_backups(parent="projects/{project_id}/locations/-")`.

#### Option B: `gcloud` CLI (Terminal / Coding Harnesses)

1. **Instances Discovery**:

```bash CLOUDSDK_METRICS_ENVIRONMENT="gcs-skills gcs-skills/1.0 (skill:google-cloud-filestore-auditing)" \ gcloud filestore instances list --project="{project_id}" --format="json" ``` 2. **Backups Discovery**:

```bash CLOUDSDK_METRICS_ENVIRONMENT="gcs-skills gcs-skills/1.0 (skill:google-cloud-filestore-auditing)" \ gcloud filestore backups list --project="{project_id}" --format="json" ```

*(Note: Do NOT pass `--location=-` to `gcloud filestore backups list`; omitting the flag queries all regions across the project automatically).*

#### Option C: GCP REST API (`call_gcp_api` in Gemini Enterprise File Agent / curl)

1. **Instances**: `GET https://file.googleapis.com/v1/projects/{project_id}/locations/-/instances` 2. **Backups**: `GET https://file.googleapis.com/v1/projects/{project_id}/locations/-/backups`

#### Option D: Standalone Script Runner

For environments with standard python3 execution enabled, the agent may invoke the portable script: `python3 scripts/filestore_audit.py --project="{project_id}" [--format=markdown|json]`

--------------------------------------------------------------------------------

### 2. Audit Evaluation & Vector Checks

For each discovered instance, evaluate three audit vectors:

#### Vector 1: Disaster Recovery & Backup Protection

- Match backups to instances using full canonical resource URIs (`backup.sourceInstance == instance.name`). - **Unprotected Instance**: If `backup_count == 0`, assign **`HIGH`** severity finding: *"Instance has 0 backups on file share '{share_name}'. Disaster recovery is not configured."* - **Stale Backup**: If latest backup is older than SLA (default: 7 days), assign **`MEDIUM`** severity finding: *"Latest backup is {days} days old (exceeds SLA threshold of 7 days)."* - Refer to `references/backup-dr-governance.md` for backup retention and SLA rules.

#### Vector 2: Security & Access Governance

- Inspect `fileShares[0].nfsExportOptions`: - **Open Network Exposure**: If `0.0.0.0/0`, `0.0.0.0`, or `::/0` is present in `ipRanges`, assign **`CRITICAL`** severity (if `accessMode: READ_WRITE`) or **`HIGH`** severity (if `READ_ONLY`): *"Export rule exposes share to 0.0.0.0/0 with accessMode='{access_mode}'."* - **Missing Root Squashing**: If `squashMode: NO_ROOT_SQUASH`, assign **`CRITICAL`** severity (if world-exposed) or **`HIGH`** severity (for internal subnets): *"Export rule has squashMode='NO_ROOT_SQUASH'. Remote root clients retain superuser UID 0 privileges."* - **Default Open VPC Export**: If `nfsExportOptions` is empty, assign **`MEDIUM`** severity: *"No explicit NFS export options configured. Share defaults to open client access within VPC with NO_ROOT_SQUASH."* - Refer to `references/security-access-governance.md` for export option configurations.

#### Vector 3: Zone Isolation & Reliability Compliance

- **Physical Zone Isolation (PZI)**: If `satisfiesPzi: false`, assign **`MEDIUM`** severity: *"Instance does not satisfy Physical Zone Isolation (PZI)."* - **Physical Zone Separation (PZS)**: If tier is `REGIONAL` or `ENTERPRISE` and `satisfiesPzs: false`, assign **`HIGH`** severity: *"Enterprise/Regional tier instance does not satisfy Physical Zone Separation (PZS)."* - **Performance Limits**: Extract `performanceLimits.maxWriteIops` and `performanceLimits.maxReadThroughputBps` (convert to MB/s). - Refer to `references/zone-isolation-pzi-pzs.md` for datacenter failure domain isolation standards.

--------------------------------------------------------------------------------

### 3. Executive Posture Scoring & Grading

Calculate fleet posture grade as defined in `references/audit-rules-matrix.md`:

- **Grade F (🔴 CRITICAL RISK)**: $\ge 1$ `CRITICAL` findings. - **Grade C (🟠 ELEVATED RISK)**: 0 Critical, but $\ge 1$ `HIGH` findings. - **Grade B (🟡 MODERATE)**: 0 Critical/High, but $\ge 1$ `MEDIUM` findings. - **Grade A (🟢 HEALTHY)**: 0 findings across all vectors.

Calculate `Backup Protection Rate %`: $$\text{Protection Rate} = \frac{\text{Total Instances} - \text{Unprotected Instances}}{\text{Total Instances}} \times 100$$

--------------------------------------------------------------------------------

### 4. Required Output Format

**Every audit report response MUST include the following structured sections in Markdown:**

#### 1. Executive Posture Scorecard

```markdown ## Executive Posture Scorecard: `{project_id}`

| Metric | Status | Details | | :--- | :--- | :--- | | **Overall Health Posture** | **[🔴 CRITICAL RISK / 🟠 ELEVATED RISK / 🟡 MODERATE / 🟢 HEALTHY]** | [Grade F / C / B / A] | | **Instances Audited** | `[count]` | Total Filestore instances evaluated | | **Backup Protection Rate** | **[pct]%** | `[protected]/[total]` instances have active backups | | **Critical & High Security Findings** | `[crit] Critical, [high] High` | Open exports (0.0.0.0/0) or NO_ROOT_SQUASH | | **PZI Isolation Compliance** | `[pzi_count]/[total]` | Physical Zone Isolation adherence | ```

#### 2. Priority Findings & Remediation Matrix

List findings ranked by severity (`CRITICAL` $\to$ `HIGH` $\to$ `MEDIUM` $\to$ `LOW`):

```markdown ## Priority Findings & Remediation Matrix

| Severity | Instance ID | Category | Finding Description | Remediation Plan | | :--- | :--- | :--- | :--- | :--- | | 🚨 **CRITICAL** | `[instance]` | Security | [Description] | [Remediation] | | ⚠️ **HIGH** | `[instance]` | Disaster Recovery | [Description] | [Remediation] | ```

#### 3. Filestore Instance Inventory & Compliance Status

```markdown ## Filestore Instance Inventory & Compliance Status

| Instance ID | Location | Tier | Capacity | Reserved CIDR | Write IOPS | Throughput | PZI | PZS | Backups | Latest Backup | | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | | `[id]` | `[loc]` | `[tier]` | `[cap] GiB` | `[cidr]` | `[iops]` | `[tp] MB/s` | ✅ Yes / ❌ No | ✅ Yes / ❌ No / N/A | 🔴 0 / ✅ `[n]` | `[date]` | ```

#### 4. Automated Remediation & Mandatory Confirmation Gate

If any unprotected instances are discovered, display attributed backup creation commands and conclude with an explicit confirmation request:

```bash CLOUDSDK_METRICS_ENVIRONMENT="gcs-skills gcs-skills/1.0 (skill:google-cloud-filestore-auditing)" \ gcloud filestore backups create [INSTANCE]-backup-[DATE] \ --project=[PROJECT_ID] \ --instance=[INSTANCE] \ --file-share=[SHARE] \ [--instance-zone=[ZONE] | --instance-location=[REGION]] \ --region=[BACKUP_REGION] ```

> **Confirmation Required**: Would you like me to proceed with creating baseline > backups for the unprotected instances? Please confirm to proceed.

**CRITICAL RATIONALE**: Do NOT execute backup creation without explicit user confirmation. Confirmation is strictly required before executing any mutation or backup creation commands in order to prevent unintended operational disruption, unwanted resource allocation, or unexpected backup storage billing.

--------------------------------------------------------------------------------

## Reference Directory

For progressive disclosure and deep architectural guidance, consult the following references:

- [Security & Access Governance](references/security-access-governance.md) - [Disaster Recovery & Backup Governance](references/backup-dr-governance.md) - [Zone Isolation (PZI) & Reliability (PZS)](references/zone-isolation-pzi-pzs.md) - [Audit Rules & Scoring Matrix](references/audit-rules-matrix.md) - [Standalone Python Audit Script](scripts/filestore_audit.py)

Intended uses

  • Use Google Cloud Filestore Auditing when this documented workflow matches the task.

Implementation notes

  • Review the source instructions and adapt inputs before running the workflow.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.