| name | sybra-monitor |
| description | Deprecated. Monitor now runs in-process inside Sybra. Use `sybra-cli monitor scan` for an ad-hoc detector pass or read the `monitor:report` event from the GUI. |
| allowed-tools | Bash, Read |
| user-invocable | true |
Sybra Monitor (legacy skill stub)
The monitor loop previously driven by /loop 5m /sybra-monitor now runs
inside the Sybra Go backend (internal/monitor/). It ticks every 5 min,
detects anomalies, applies idempotent remediations, and spawns focused
headless Claude agents for anything that needs LLM judgment.
This skill no longer runs in a loop. Do not register a cron or /loop
for it.
Read-only ad-hoc pass
sybra-cli monitor scan
sybra-cli monitor scan --json
The --json output matches monitor.Report in internal/monitor/report.go
and is emitted over the Wails monitor:report event on every tick.
What moved to Go
- Board + audit snapshot, dwell check, threshold detection —
internal/monitor/detector.go.
lost_agent reset + untriaged status reason — internal/monitor/remediator.go.- Issue dedup + creation via
gh — internal/monitor/issuesink.go.
- LLM dispatch (pr_gap, stuck_human_blocked, failure_spike, bottleneck) —
internal/monitor/dispatcher.go via agent.Manager.Run.
- Heartbeat — removed. The in-process service is its own liveness signal.
If something in the loop is broken, start at internal/monitor/service.go
and follow the references.
