fork-sync-review

name	fork-sync-review
description	Review fork divergence, sync origin/main with upstream/main, keep this fork aligned as a deployment wrapper around the published upstream nousresearch/hermes-agent image, preserve data/.env and Postgres state, and produce a smoke-test plan for deployment updates.
argument-hint	Describe the fork sync, merge, or image migration question
user-invocable	true

Fork Sync Review

Use this skill when the task is to inspect how far this fork has drifted from upstream, decide whether deployment changes should stay fork-local, or keep the fork aligned as a deployment wrapper around the published upstream image without losing runtime state.

Procedure

1. Inspect remotes and divergence with git remote -v, git fetch upstream --prune, git fetch origin --prune, and git rev-list --left-right --count origin/main...upstream/main.
2. Review fork-only deployment files before changing the running stack: docker-compose.yml, INSTALL.md, docker/deploy.sh, and docker/hermes-config.yaml. Read docker/entrypoint.sh when startup behavior or entrypoint drift is part of the question.
3. Check the currently running containers and images with docker compose ps and compare them to the documented upstream image flow in website/docs/user-guide/docker.md and docker publish workflow.
4. Preserve local state. Never delete data/.env, data/config.yaml, data/memories/, data/sessions/, or the PostgreSQL volume unless the user explicitly asks.
5. Prefer nousresearch/hermes-agent:latest for upstream image pulls unless a different official tag has been explicitly verified. Do not assume a GHCR package exists without checking.
6. After any deployment edit, validate with docker compose config, then use docker compose -f docker-compose.upstream.yml up -d --pull always --force-recreate --remove-orphans for upstream-image stacks, followed by health checks on the dashboard and API endpoints.

Output Expectations

• Report fork divergence as ahead/behind counts.
• Separate fork-only deployment customizations from upstream changes that can be adopted safely.
• Call out anything that would change credentials, mounted config, or persistent data before doing it.
• End with the concrete commands needed to smoke test the updated stack.