Monitor

Monitor

Strategy

1. Select preset (determines data source and key metrics)
2. Gather health data from the source
3. Compare against baseline/thresholds
4. Rate: healthy / degraded / critical
5. Format report
6. Alert if critical
7. If --recurring: schedule next run

Exit: report delivered. For recurring: runs until cancelled.

Report requirements (every report MUST include)

• Specific metric values: exact numbers, not just "high" or "degraded"
• Trend direction: is each metric improving, stable, or worsening vs baseline?
• Actionable items: concrete next steps ranked by urgency, not just observations
• Anomaly callouts: flag anything outside expected range with the specific threshold breached

> Note: Placeholders like {user_question} in Agent prompts are filled by you (Claude) > from the current task context. They are not template variables — read the user input, > gather the relevant context, and substitute before spawning the agent.

Agents

GATHER phase

Agent(subagent_type="Explore", model="haiku", prompt="""
Gather health data for: {target}
Preset: {preset}

Data sources to check:
{preset_data_sources}

Key metrics to collect:
{preset_metrics}

Output: raw metrics with timestamps.
""")

ASSESS + REPORT phase

Agent(model="sonnet", prompt="""
Health data:
{gathered_metrics}

Baseline (last {baseline_days} days):
{baseline_data}

1. Compare current vs baseline
2. Flag anomalies (>2 stddev from baseline)
3. Rate overall: HEALTHY / DEGRADED / CRITICAL
4. Format as a concise health report
""")

Presets

| Preset | Sources | Metrics | |---|---|---| | --service NAME | Observability platform, tracing | Latency p50/p99, error rate, throughput, instance count | | --pipeline NAME | Pipeline orchestrator | Success rate, SLA compliance, last failure | | --ci | CI/CD platform | Build success rate, flaky test %, avg build time | | --deploy APP | Deployment platform | Deploy state, canary score, pending constraints | | --ml FLOW | ML platform | Run status, latest metrics, accuracy trend | | --code REPO | Git history | TODO/FIXME count, test coverage, PR merge rate | | --deps REPO | pip/npm audit, CVE DBs | Outdated count, critical CVEs, last updated | | --docs | Documentation platform | Pages not updated in 90d, broken links | | --team NAME | GitHub, Slack, Jira | PR velocity, open issues, Slack activity | | --oncall | PagerDuty, incident log | Open incidents, MTTR, handoff notes |

Cross-provider review

When cross-provider tools are available, run verification on a non-Claude model in parallel for maximum blind-spot diversity.

Flags

• --recurring=INTERVAL — repeat on schedule (daily, hourly, etc.)
• --alert=CHANNEL — notification channel for critical alerts
• --baseline=N — days of baseline data (default: 14)

Examples

/monitor --service my-api
/monitor --pipeline etl-daily --recurring=daily
/monitor --ci --alert=#builds
/monitor --team platform-eng
/monitor --code --baseline=30
/monitor --oncall

Execution routing (sagaflow-first)

Sagaflow is the default execution path. The in-session workflow above is the FALLBACK, used only when the sagaflow worker is confirmed unavailable.

Routing sequence (mandatory before any in-session work):

1. If one-shot health check (not --recurring) → produce report directly in-session. Do NOT route to sagaflow.
2. Run sagaflow doctor
3. If healthy → launch via sagaflow below. Stop. Do not run in-session.
4. If unhealthy → log SAGAFLOW_UNAVAILABLE, proceed with in-session fallback.

Launch command:

Bash(
  run_in_background=true,
  command="sagaflow launch monitor --arg target='' --arg preset=auto --arg recurring=false --await"
)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: npow
• Source: npow/claude-skills
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.