Audits Google Cloud Filestore instances 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 ROOT_SQUASH), 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.
Permissions
Files
SKILL.md
google-cloud-filestore-auditing
Audits Google Cloud Filestore instances 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 ROOT_SQUASH), 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.
metadata.category
Storage
Google Cloud Filestore Auditing Skill
This skill enables autonomous agents to audit, evaluate, and report the disaster
recovery, security access governance, and architectural reliability posture of
Google Cloud Filestore fleets across GCP projects.
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.
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:
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.
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:
First, check if Filestore MCP tools (list_instances, list_backups)
are available and query them directly (these are API invocations, not
shell command executions).
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).
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)
CLOUDSDK_METRICS_ENVIRONMENT="gcs-skills gcs-skills/1.0 (skill:google-cloud-filestore-auditing)" \
gcloud filestore instances list --project="{project_id}" --format="json"
Backups Discovery:
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)
Instances: GET https://file.googleapis.com/v1/projects/{project_id}/locations/-/instances
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.
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:
