Continuity runbook: Deploy production

Task

A second maintainer understands how production ships (Actions → Heroku force-push + Alembic release: guardrail) and can dispatch a deploy only when explicitly told.

Success criteria

• Can explain current Heroku release vs origin/main and the last “Deploy to OPENCREORG” run
• Knows make alembic-guardrail / scripts/check_alembic_revision_guardrail.py and that a guardrail fail leaves the old slug
• Will not git push --force GitHub main or laptop-push to Heroku unless following the documented path after DEPLOY_OPENCREORG
• Runbook docs/continuity/runbooks/deploy-production.md (skill continuity-deploy-production)

Context

• .github/workflows/deploy.yml, Procfile (release, web, worker)
• Staging is a different app/workflow (tmp-cre, deploy-staging.yml)

Constraints

• existing maintainers only
• Default mode is read-only until the human says DEPLOY_OPENCREORG
• Never print DATABASE_URL when running the guardrail against prod

Agent prompt

See Agent prompt in docs/continuity/runbooks/deploy-production.md.

Start with docs/continuity/runbooks/deploy-production.md, then read .github/workflows/deploy.yml, Procfile, and scripts/check_alembic_revision_guardrail.py. Verify the runbook covers the listed production-release checks, the Alembic guardrail commands, the read-only default, and the explicit DEPLOY_OPENCREORG requirement without exposing DATABASE_URL.