See what Watchdog notices.

Coverage preview

Lifecycle monitoring

Configured, awaiting evidenceConfiguration: Start, completion and reported failure monitoring enabledObserved evidence: No run evidence receivedResult: Awaiting an authenticated run start.A silent process crash is inferred from a missed deadline, not directly observed. Missing optional instrumentation does not fail a lifecycle-only run.

Runtime deadline

Configured, awaiting evidenceConfiguration: 1200 seconds maximum runtimeObserved evidence: No explicit run start receivedResult: Awaiting a run start to anchor the deadline.Uses the start event timestamp from accepted lifecycle telemetry. A runtime deadline does not prove business success.

Expected schedule

Not configuredConfiguration: No expected scheduleObserved evidence: Schedule-check history is shown in missed-execution incidents and SLOs, not inferred from run eventsResult: No missed-start check requested.The independent background checker detects missed starts; it does not launch the job.

Progress monitoring

Not configuredConfiguration: Milestone timeout disabledObserved evidence: 0 explicit progress events receivedResult: Optional; lifecycle monitoring works without milestones.When enabled, the timer starts at run start even before the first milestone. Tools, heartbeats and model responses are not business progress.

Tool-loop detection

Configured, awaiting evidenceConfiguration: Repeated-cycle threshold: 4; requires fingerprinted tool completionsObserved evidence: 0 of 0 observed tool completions have fingerprintsResult: Repeated-cycle check has no fingerprinted tool evidence.A wrapper or adapter observes only its instrumented calls. One event never proves complete tool coverage; adapters may omit tool failures or tools outside their hooks. Explicit progress resets the cycle window.

Cost monitoring

Configured, awaiting evidenceConfiguration: $3 limit on reported cost deltasObserved evidence: No monetary cost samples received; spend is unknownResult: Budget check cannot establish spend without cost data.An explicit measured zero is a sample; missing data is not $0. Token counts do not calculate dollars. Separate usage deltas cannot prove which model events or other charges they cover.

Outcome verification

Not configuredConfiguration: No required outcomes or external verifierObserved evidence: 0 self-reported outcome events; external verdict: not receivedResult: Not independently verified. Normal completion does not prove a business result.The external callback proves only what that customer-supplied check actually validates. Application-reported outcomes remain self-reported.

Recovery / model fallback

Not configuredConfiguration: Automatic cancellation and retry disabled; no control endpoint or fallback modelObserved evidence: No runtime command evidenceResult: Optional; no recovery requested.The customer runtime must cancel or launch retries and select an allowed model. A configured name or endpoint is not verified automatic recovery.

Configuration preview only. Saved settings do not establish observed coverage or a successful result. Verify a fresh run after saving changes.