Problem

020_state_machine:latest (and other tutorial-agent images) have been frozen at a December 2025 build and re-published unchanged on every run — which is why the merged mcp<2 pin never reached the image and scale-agentex's 020-state-machine integration test keeps failing on the mcp 2.0.0 McpError→MCPError crash.

Root cause — the publish path re-pushes a stale image, not the one it just built:

# build step: multi-platform, NO --push
agentex agents build ... --platforms linux/amd64,linux/arm64
# validate step:
docker run "$FULL_IMAGE" ...     # pulls the EXISTING :latest from ghcr
# push step:
docker push "$FULL_IMAGE"        # re-pushes that pulled (stale) image

A multi-platform buildx build cannot be loaded into the local Docker image store, so the build step never creates a local :latest. The validation step's docker run "$FULL_IMAGE" then pulls the existing (stale) :latest from the registry, and the separate docker push re-pushes that same stale image back. The freshly-built image is silently discarded — so :latest is self-perpetuating and stuck at its December digest.

This was verified end-to-end:

• A local single-arch docker build from the pinned source produces a fresh image with the pin.
• CI (multi-arch, build-then-docker push) re-pushed the identical December digest (1f90…, Created 2025-12-30, no pin) even with --no-cache and SDK 0.22.0.

Fix

• On the publish path, pass --push to agentex agents build so buildx sends the freshly-built multi-arch image straight to the registry.
• Drop the now-redundant separate docker push step. Validation still runs and now exercises the freshly-pushed image.
• Revert the CLI build-cache default back to True (keeping the --cache/--no-cache flag). The staleness was never a layer-cache problem, so forcing --no-cache globally only slowed builds.

Effect

The next publish (push to main under examples/tutorials/**, or a rebuild_all dispatch) will push the freshly-built, pinned 020_state_machine:latest, turning the scale-agentex integration check green — and fixes the same silent-stale-publish for every tutorial agent.

Note on the earlier PR

This supersedes the root-cause theory behind the earlier no-cache change (#476): --no-cache was confirmed to run in CI yet the image stayed stale, which is what led to finding the real build/push bug here.

🤖 Generated with Claude Code

Greptile Summary

This PR fixes a long-standing silent publish bug where multi-platform buildx builds were being silently discarded and a stale :latest re-pushed in its place. The fix changes the publish path to push directly from buildx to an immutable SHA-tagged candidate, validate that exact image, then promote it to :latest via docker buildx imagetools create — keeping :latest clean until validation passes. The cache default is also reverted to True since caching was never the root cause of staleness.

• Publish path overhauled: build step now passes --push so buildx sends the fresh multi-arch manifest straight to the SHA tag; a separate "promote" step then points :latest at that validated digest with docker buildx imagetools create, preserving the multi-arch manifest list without rebuilding.
• Validation now exercises the live SHA-tagged image: because multi-platform images cannot be loaded into the local Docker store, validation pulls the freshly-pushed registry image — an inherent constraint of multi-arch buildx noted in the PR description.
• Cache default reverted: cache: bool = True in both agents.py and agent_handlers.py; --no-cache remains available but is no longer the default since layer-cache staleness was not the root cause.

Confidence Score: 5/5

Safe to merge — the change correctly routes the publish path through buildx --push, validates the live SHA-tagged image, and only then promotes :latest.

The three changed files are tightly scoped: the workflow correctly pushes from buildx to an immutable SHA tag, validates that exact image, and promotes :latest via a registry-side manifest copy. The Python cache-default reversal is a one-line change with no side effects. The only finding is a dead env variable that is exported but never read, which has no functional impact.

Files Needing Attention: No files require special attention beyond a quick read of the SHOULD_PUSH vs PROMOTE_LATEST interplay in the workflow.

Important Files Changed

Reviews (2): Last reviewed commit: "validate before promoting :latest (candi..." | Re-trigger Greptile