Document metadata
- Status
- Maintained
- Approval
- Approved
- Version
- 1.0
- Classification
- PUBLIC
- Owner
- Lightning IT Documentation Maintainers
- Approver
- Lightning IT Security and Compliance Maintainers
- Audience
- security engineers, automation operators
- Last reviewed
- Next review
- (Semiannual)
ModuLix security
Automation concentrates privilege: content can translate one authorization into changes across many targets. Treat content, dependencies, inputs, runtime, and evidence as separate trust boundaries.
Trust-boundary review
| Boundary | Main risks | Required decision |
|---|---|---|
| Source and release | Tampering, dependency substitution, unreviewed change | Pin an immutable reviewed artifact and verify available provenance |
| Inventory and variables | Scope expansion, sensitive data disclosure, unsafe defaults | Validate schema, ownership, classification, and exact target limit |
| Secret resolution | Exposure in source, command history, process output, or logs | Supply secrets through an approved runtime channel and suppress disclosure |
| Controller and execution context | Excess privilege, mutable dependencies, credential persistence | Isolate the runtime and grant only the access required for the bounded task |
| Managed target | Destructive change, partial convergence, unintended reach | Preflight, authorize, limit concurrency, verify, and define recovery |
| Evidence | Credentials or environment data retained too broadly | Redact, classify, restrict access, and apply retention |
Content review
Review changes for command construction, template injection, unsafe file modes, unbounded loops, implicit downloads, certificate-verification bypasses, logging of sensitive variables, and behavior that crosses the role's stated scope. Dependencies deserve the same review as first-party tasks.
Secret handling
Public content may name a secret input contract, but it must not contain the value, a recoverable encrypted payload, a real secret path, or a credential-like example. Avoid secrets on command lines because shell history and process lists can expose them. Mark sensitive Ansible tasks with appropriate output controls, then verify that failure paths also remain redacted.
Recovery is not automatic
Automation rollback is safe only when the selected content explicitly supports and tests it for the relevant state. Otherwise, recovery may mean restore, repair, or a reviewed forward change. Identify the recovery owner and backup precondition before execution; do not improvise after a partial failure.
This page describes a security approach, not a claim that every component or deployment implements every control. The site-wide security model and backup and recovery guidance provide the cross-product context.