Spanda

Plugin Security

Spanda plugins run under a defense-in-depth model: manifest validation, capability review, trust tiers, signature verification, sandboxing, and audit logging.

Requirements

Control	Behavior
Signature verification	Ed25519 over canonical plugin payload
Manifest validation	Schema, plugin type, capability names
Capability review	Dangerous capabilities require --approve-dangerous
Sandboxing	WASM default; [security].sandbox = true
Trust tier	Registry tier gates install (blocked rejected)
Audit logging	Install, enable, disable, hook dispatch logged
Version compatibility	Spanda semver + API v1 enforced at install

Blocked installs

Install is rejected when:

• Plugin registry tier is blocked
• Official plugin has signed = true but signature verification fails
• Plugin requests dangerous capabilities without approval
• Plugin is incompatible with current Spanda or plugin API version

Dangerous capabilities

Require spanda plugin install --approve-dangerous:

• entity.write, device.write, filesystem.write
• network.outbound, readiness.write, recovery.write

Trust tiers

Tier	Install
official	Allowed; signature required when signed = true
verified	Allowed
community	Allowed
experimental	Allowed
deprecated	Allowed (discouraged)
blocked	Denied

Relationship to packages

Plugin security is orthogonal to package trust. Packages continue to use spanda.toml and registry provenance unchanged.

This site is open source. Improve this page.