Description

Description

The monitoring page (docs/monitoring/monitor-sandbox-activity.md) is the primary
entry point for users debugging sandbox issues, but it omits three capabilities
that already ship in v0.0.43:

1. Inference health third state: the "Check Sandbox Health" section (line 47)
   only documents `healthy` and `unreachable`, but `nemoclaw  status`
   also outputs `not probed` for compatible-* providers where the endpoint URL
   is unknown. Users on compatible endpoints see this state with no doc explanation.

2. `doctor` command: `nemoclaw  doctor [--json]` runs grouped diagnostics
   across Host / Gateway / Sandbox / Inference / Messaging / Local services.
   The monitoring page never mentions it.

3. `logs` flags and description: the page only shows `--follow`, but the CLI
   also supports `--tail ` (alias `-n`) and `--since `.
   The page describes logs as "blueprint runner and sandbox" output, but
   `--help` says "Show OpenClaw gateway logs and OpenShell audit logs" —
   the OCSF audit log source is undocumented on this page.

Device:        MacBook Pro (Apple Silicon)
OS:            macOS 15 (Darwin 25.0.0, arm64)
Architecture:  arm64
Node.js:       v22.22.1
npm:           10.9.4
Docker:        29.2.1
OpenShell CLI: 0.0.39
NemoClaw:      v0.0.43
OpenClaw:      v2026.4.24

1. Read docs/monitoring/monitor-sandbox-activity.md

2. Inference "not probed" state:
   nemoclaw my-assistant status
   → Observe: Inference: not probed (...)
   → Search the monitoring page for "not probed" — not found

3. doctor command:
   nemoclaw my-assistant doctor
   → Outputs grouped health check (Host/Gateway/Sandbox/Inference/Messaging/Local)
   → Search the monitoring page for "doctor" — not found

4. logs flags:
   nemoclaw my-assistant logs --tail 5
   nemoclaw my-assistant logs --since 1m
   → Both succeed
   → Search the monitoring page for "--tail" or "--since" — not found
   nemoclaw my-assistant logs --help
   → Shows "Show OpenClaw gateway logs and OpenShell audit logs"
   → Page says "blueprint runner and sandbox" — different description

1. "Check Sandbox Health" section documents all three inference states:
   healthy, unreachable, not probed (with explanation of when each appears)
2. A "Run Diagnostics" section mentions `nemoclaw  doctor [--json]`
3. "View Logs" section documents --tail and --since flags, and describes
   the dual log sources (gateway + OCSF audit)

1. Only "healthy" and "unreachable" documented (line 47)
2. doctor command not mentioned anywhere on the page
3. Only --follow shown; log description says "blueprint runner and sandbox"
   instead of "gateway logs and OpenShell audit logs"

$ nemoclaw my-assistant status | grep Inference
    Inference: not probed (Kimi K2.6 health requires NVIDIA_API_KEY; skipping ...)

$ nemoclaw my-assistant doctor
  Host:      [ok] CLI build / Docker / OpenShell
  Gateway:   [fail] Docker container not found
  Sandbox:   [ok] Ready
  Inference: [ok] reachable
  Summary:   attention needed (1 failed, 1 warning)

$ nemoclaw my-assistant logs --tail 5
  (returns last 5 log lines successfully)

$ nemoclaw my-assistant logs --since 1m
  Warning: log buffer contains only the last 200 lines; --since results may be incomplete.
  (returns filtered logs successfully)

$ nemoclaw my-assistant logs --help | head -3
  USAGE
    $ nemoclaw  logs [--follow] [--tail |-n ] [--since ]
  DESCRIPTION
    Show OpenClaw gateway logs and OpenShell audit logs

Bug Details

[NVB#6186562]