Audit and rework cron/launchd jobs for Docker architecture

[mriechers]

Enhancement

Context: Any scheduled jobs (cron or macOS launchd) associated with Cardigan / ai-editorial-assistant-v3 were built for the pre-Docker local architecture. They need to be audited and reworked for the containerized version.

Problem

Scheduled jobs that interact with Cardigan may:

• Reference local filesystem paths (OUTPUT/, transcripts/) that are now inside Docker volumes
• Call scripts directly that now run inside containers
• Assume the API is at localhost:8000 when it may be behind nginx or a Cloudflare Tunnel
• Interact with the MCP server via stdio when it needs to move to SSE (Refactor MCP server to work within Docker architecture #35)

Known Launch Agents (the-lodge namespace)

The following ~/Library/LaunchAgents/ may touch the Cardigan workflow — needs human audit to confirm which are relevant:

Agent	Status	Possibly Relevant?
com.the-lodge.airtable-sync	Loaded	Likely — syncs Airtable data that Cardigan reads
com.the-lodge.mcp-config-sync	Loaded (exit 127)	Likely — manages MCP server configuration
com.the-lodge.cli-agent-http	Loaded (exit 78)	Possibly — HTTP agent that may call Cardigan API
com.the-lodge.diane-listener	Loaded (exit 78)	Possibly — event listener
com.the-lodge.github-issues-sync	Loaded	Unlikely but check
com.the-lodge.agent-registrar	Loaded	Unlikely but check
com.the-lodge.log-lady-audit	Loaded	Unlikely but check

Note: No direct cardigan or editorial-assistant entries found in crontab or LaunchAgents filenames — the relevant jobs are likely embedded within the-lodge infrastructure.

Rework Options

For each relevant job, decide:

1. Move into Docker — Convert to a container-based cron (e.g., ofelia, or a dedicated cron container in docker-compose) that runs inside the Docker network
2. Keep on host, update targets — Keep the launchd agent but update it to call the Docker API endpoint instead of local scripts
3. Replace with Docker healthcheck/scheduler — Use Docker's built-in healthcheck or a worker container's internal scheduler
4. Retire — If the job's purpose is handled by the new architecture (e.g., the worker container replaces a cron-triggered script)

Acceptance Criteria

• All cron/launchd jobs touching Cardigan are identified
• Each is categorized: move to Docker, update target, or retire
• Updated jobs work correctly with the containerized architecture
• No orphaned jobs reference paths or services that no longer exist locally