VMware Policy

Disclaimer: This is a community-maintained open-source project and is not affiliated with, endorsed by, or sponsored by VMware, Inc. or Broadcom Inc. "VMware" and "vSphere" are trademarks of Broadcom. Source code is publicly auditable at github.com/vmware-skills/VMware-Policy under the MIT license.

Unified audit logging, policy enforcement, and input sanitization -- the infrastructure layer for the entire VMware MCP skill family.

Infrastructure dependency: Every other package in the VMware skill family depends on vmware-policy. It is auto-installed and provides the @vmware_tool decorator, sanitize(), and the shared audit database. Family: vmware-aiops (VM lifecycle), vmware-monitor (read-only monitoring), vmware-storage (iSCSI/vSAN), vmware-vks (Tanzu Kubernetes), vmware-nsx (NSX networking), vmware-nsx-security (DFW/firewall), vmware-aria (metrics/alerts/capacity), vmware-avi (AVI/ALB/AKO). | vmware-pilot (workflow orchestration)

What This Skill Does

Quick Install

uv tool install vmware-policy==1.24.2
vmware-audit stats          # verify installation

vmware-policy is automatically installed as a dependency of all VMware skills. Manual install is only needed for standalone audit querying.

When to Use This Skill

• Query the unified audit trail across all VMware skills
• View denied operations and policy violations
• Check audit statistics (by skill, by status, by time range)
• Export audit logs as JSON for external analysis
• Configure deny rules or maintenance windows
• Integrate the @vmware_tool decorator into a new VMware skill

This skill is auto-loaded as a dependency -- you do not need to invoke it directly. It activates when:

• Any VMware skill tool function is called (via @vmware_tool decorator)
• User asks about audit logs, denied operations, or policy rules
• User runs vmware-audit CLI commands

Related Skills -- Skill Routing

Common Workflows

Query Recent Audit Activity

1. View last 20 audit entries: vmware-audit log --last 20
2. Filter by skill: vmware-audit log --skill nsx --last 50 (the skill column holds the short name: nsx, aiops, nsx_security, ...)
3. Check denied operations: vmware-audit log --status denied --since 2026-03-28
4. View aggregate stats: vmware-audit stats --days 7

Set Up Policy Rules for Production

1. Copy default rules: cp $(python -c "import vmware_policy; print(vmware_policy.__file__.replace('__init__.py','rules_default.yaml'))") ~/.vmware/rules.yaml (the shipped baseline denies nothing -- every rule in it is commented out)
2. Edit ~/.vmware/rules.yaml -- add deny rules for production:

   deny:
     - name: no-delete-in-prod
       operations: ["delete_*", "remove_cluster_host"]
       environments: ["production"]
       reason: "Destructive operations blocked in production"
   maintenance_window:
     start: "22:00"
     end: "06:00"
   

3. Rules hot-reload automatically -- no restart needed
4. Confirm the rules loaded: vmware-audit policy prints the rule source; vmware-audit policy --operation delete_vm --skill aiops --env production --risk critical shows what happens to one call
5. Verify: vmware-audit log --status denied to see blocked operations

Export Audit Logs for Compliance

1. Export all logs as JSON: vmware-audit export --format json > audit-export.json
2. Filter by skill: vmware-audit export --skill aiops --since 2026-01-01
3. Import into your SIEM or compliance tool

Usage Mode

CLI Quick Reference

# View recent audit entries
vmware-audit log --last 20
vmware-audit log --skill nsx --status denied
vmware-audit log --since 2026-03-28 --tool delete_segment

# Which rules are in force (user file, shipped baseline, or unreadable)
vmware-audit policy

# Export for compliance
vmware-audit export --format json > audit.json
vmware-audit export --skill aiops --since 2026-01-01

# Aggregate statistics
vmware-audit stats --days 7
vmware-audit stats --days 30

Full CLI reference: see references/cli-reference.md

Python API

from vmware_policy import vmware_tool, sanitize

# Wrap every MCP tool function
@vmware_tool(risk_level="high", sensitive_params=["password"])
def delete_segment(name: str, env: str = "") -> dict:
    ...

# Sanitize untrusted API responses before returning to LLM
clean_text = sanitize(api_response_text, max_len=500)

MCP Tools (0)

vmware-policy does not expose MCP tools. It is a Python library and CLI consumed by other VMware skills.

Agent Detection (Transparency Note)

The detect_agent() function in audit.py checks the following environment variables to identify which AI agent is calling the tools. This is read-only inspection for audit logging purposes — no credentials are extracted or stored:

Only the presence of these variables is checked; their values are never read into the audit row, and no API-key variable is inspected. The agent column stores only the inferred name above.

Policy Defaults and Failure Behavior

Read/write authorization belongs to the vCenter/NSX account's RBAC; this engine only applies the operator's own optional deny rules and maintenance window. vmware-audit policy reports which of four rule sources is in force:

A malformed maintenance_window blocks high/critical operations. PyYAML is a declared dependency; if it is missing, the engine cannot start and every decorated tool call fails with an error (audited as error) -- nothing is silently allowed. ~/.vmware is $OPS_HOME when that variable is set.

VMWARE_POLICY_DISABLED=1 is an operator escape hatch. Read from the environment of the process running the skill (the MCP host or the CLI shell), it skips all policy evaluation -- deny rules, maintenance window, and the fail-closed state above. It does not disable auditing: a warning naming the operation and parameter names is logged, and rows from both surfaces -- MCP tools (@vmware_tool) and CLI commands wrapped by @guarded -- are recorded with a _bypassed status suffix (ok_bypassed, error_bypassed). Restrict who can set environment variables for the MCP host process, and do not leave the variable set.

Troubleshooting

"Cannot initialize audit DB" warning

The audit database directory ~/.vmware/ must be writable. Create it manually: mkdir -p ~/.vmware && chmod 700 ~/.vmware.

Policy rules not taking effect

Rules are loaded from ~/.vmware/rules.yaml ($OPS_HOME/rules.yaml if set). Run vmware-audit policy: packaged-default means your file was not found, user-unreadable means it failed to load and every operation is being denied. The PolicyEngine hot-reloads on file change -- no restart needed.

Audit log growing too large

The AuditEngine automatically rotates at 100MB, keeping the 5 most recent archives. For manual cleanup: ls ~/.vmware/audit.*.db to see archives.

"PolicyDenied" exception in skill

A deny rule in ~/.vmware/rules.yaml matched, a closed maintenance window blocked a high/critical operation, or (rule rules_unreadable) the rules file failed to load. Check vmware-audit log --status denied for the rule name and reason. Fix the rule or file rather than bypassing; VMWARE_POLICY_DISABLED=1 removes all policy for that process (see Policy Defaults above).

Decorator not detecting skill name

The @vmware_tool decorator infers the skill name from the module path (e.g., vmware_aiops.ops.vm_lifecycle -> aiops). If the module does not follow the vmware_<skill> convention, the skill is logged as "unknown".

SQLite "database is locked" error

Multiple concurrent skill processes can write to the same audit.db via WAL mode. If locks persist beyond 5 seconds, check for zombie processes holding the database file.

Setup

uv tool install vmware-policy==1.24.2
mkdir -p ~/.vmware

vmware-policy is auto-installed as a dependency of all VMware skills. The ~/.vmware/ directory is created automatically on first audit write.

Full setup guide, security details, and integration instructions: see references/setup-guide.md

Security

• Source Code: github.com/vmware-skills/VMware-Policy
• Config File Contents: ~/.vmware/rules.yaml contains only rule definitions, no credentials
• Webhook Data Scope: N/A -- vmware-policy does not send data externally
• TLS Verification: N/A -- vmware-policy does not make network connections
• Prompt Injection Protection: sanitize() strips C0/C1 control characters and Unicode format characters (zero-width, bidi overrides), then truncates (default 500 chars)
• Least Privilege: Audit database is local-only (~/.vmware/audit.db), no network exposure; the directory is set to 0700 and the database files to 0600 (best-effort)
• Audit data is sensitive: each row records tool name, parameters, result, status, OS user, inferred agent, and timestamps. Parameters named in sensitive_params are stored as ***; results from tools declaring sensitive_result=True are replaced wholesale; credential-named keys (password, token, kubeconfig, ...) are redacted automatically in both parameters and results, on the MCP and CLI surfaces alike, as is credential-shaped text in error messages. A credential under a key not on that list still needs sensitive_params. Treat audit.db and its rotated archives as confidential.
• Best-effort audit: if the database cannot be written, the call still proceeds and a warning is logged
• Policy bypass: VMWARE_POLICY_DISABLED=1 disables policy checks, not auditing -- restrict who can set it (see Policy Defaults above)

License

MIT -- github.com/vmware-skills/VMware-Policy