dvcrn Agent Dashboard Sync
Operate dashboard data sync as a no-LLM pipeline.
Hard Rules
- Keep high-frequency state out of Git commits.
- Use Cloudflare Worker + KV for runtime sync.
- Run collector from local cron (
*/2 * * * *) and do not call LLM in collector path. - Never commit or print production tokens/secrets in files, logs, or screenshots.
- Do not publish absolute paths in skill docs; use relative paths or placeholders (
<PROJECTS_ROOT>,<SHARED_ROOT>).
Scope Boundary
- This skill owns: collector, Worker ingest/read API, KV schema, cron deployment, dashboard data source wiring.
- This skill does not own: cross-agent protocol, constitution governance, Discord routing rules.
KV Data Contract (v1)
fleet:registryfleet:heartbeat:<agent_id>fleet:cron:<agent_id>fleet:runtime:<agent_id>fleet:events:recentfleet:updated_at
See references/schema.md for payload shape.
Minimal Rollout
- Deploy Worker + KV namespace.
- Configure dashboard env to
cloudflaremode. - Install collector cron on each node with unique
AGENT_ID. - Verify
/health, then/fleet, then dashboard UI.
Security Checklist
- Store
INGEST_TOKENandREAD_TOKENas worker secrets. - Keep dashboard read token server-side (
DASHBOARD_READ_TOKEN), never client-exposed. - Keep
NEXT_PUBLIC_*vars non-sensitive only. - Redact tokens before sharing commands/logs.
Runbook Links
- Worker setup and command sequence:
references/worker-setup.md - Collector and crontab setup:
references/collector-cron.md - Env variable matrix:
references/env-matrix.md - Data schema reference:
references/schema.md - Security policy for all agents:
references/security-rules.md