stella self update — Command Guide

Audience: Operators and air-gapped administrators who inspect local product update manifests. Scope: Offline DSSE manifest verification and read-only update planning.

Manifest-check source verified against commit e25dab6fa702d22df9ddcf0445310645cbbc2c11 (PCD-389); trust-root resolution re-verified 2026-10-27 for SPRINT_20261027_005 VTR-2. Re-verify with pwsh ./tools/scripts/test-targeted-xunit.ps1 -Project src/Cli/__Tests/StellaOps.Cli.Tests/StellaOps.Cli.Tests.csproj -Class '*AgentSelfUpdateVerifyTests' -BuildProjectReferences -MaxThreads 1.

self update --check and self update --dry-run verify a local DSSE product update manifest against a release trust root, enforce lifecycle policy and print the verified manifest’s version, image digests and fixed CVEs. Both modes require real manifest and trust inputs; verification always runs, whether or not --verify is supplied.

Package application is unsupported. Invocations without --check or --dry-run, including --verify alone, fail nonzero before download, package-signature, rollback, apply or health claims. --force also fails; it does not enable application or a maintenance-window override. No agent rollback command is registered.

Commands

stella self update --check --manifest product-update-manifest.dsse.json --trust-root release-trust-root.pem
stella self update --dry-run --manifest product-update-manifest.dsse.json --trust-root release-trust-root.pem
stella self update --verify --dry-run --allow-eol --manifest product-update-manifest.dsse.json --trust-root release-trust-root.pem

Verification

Offline Use

Air-gapped operators pass an Offline Update Kit manifest through --manifest. The pinned vendor set verifies it without further input. To verify against another root, pass it through --trust-root; obtain that root independently of the kit it verifies and keep it outside the kit: a public key copied out of the same kit proves only that the kit is self-consistent, not that the vendor signed it. Inline --trusted-public-key-pem is supported for test harnesses and one-off recovery, but file-backed trust roots are preferred for repeatable scripts.