Continuity runbook: Production health

Task

A second maintainer can tell whether production is healthy using HTTP-only checks: gap-analysis cache, chatbot shell, REST lists — without computing GA on Heroku or sending paid chat prompts.

Success criteria

• make monitor-ga-health-prod, make verify-ga-complete-prod, make monitor-chatbot-health-prod run; exit codes recorded
• /rest/v1/standards and /rest/v1/ga_standards non-empty
• JSON written under tmp/ (not committed)
• Failures name the next continuity runbook (restart / rollback / gap-analysis-ops / catalog)
• Runbook docs/continuity/runbooks/production-health.md (skill continuity-production-health)

Context

• Same checks as the weekly prompt in AGENTS.md
• Unauthenticated /completion must be 401, not 404/500/503
• Prod GA is cache-only; miss → 404, never compute on opencreorg

Constraints

• existing maintainers only
• No authenticated chatbot prompts
• No heroku config dumps

Agent prompt

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

Read AGENTS.md and docs/continuity/runbooks/production-health.md, including the Agent prompt. Run the listed make targets and HTTP checks without authenticated chatbot prompts or Heroku config dumps, recording exit codes and writing JSON under tmp/. Done means the runbook exists, both REST lists are non-empty, /completion returns 401 unauthenticated, and failures name the next continuity runbook.