LDCon diagnostic task instructions, revision 1.0

Objective: assess this laser over the supplied time window using the common instructions and current metric/capability manifest. This task grants evidence collection and analysis only. It does not authorize tuning, lock stop/start, configuration changes, deployment or release of a reservation. Recommend an adjustment separately if warranted; the coordinator must issue its own scoped task.

1. Establish the current control job, owner generation, target, operating activity and pending work. Separate current state from historical failures and intentional stops. Use current owner inspection when the task requires physical-state verification; a recorded job snapshot is not independent owner evidence.
2. Check acquisition times, source validity, units, completeness and the requested window. Retrieve missing pages or artifacts through the advertised read routes within the query budget. Record backend ingestion delay, truncation and unresolved gaps. Do not call an incomplete excerpt a complete review.
3. Inspect all applicable required metric groups: frequency error and stability, effective acquisition cadence and jitter, output/integrator headroom, slow-loop drift, recovery events and configured optical quality. Include fresh dither and position-free secondary-peak/width evidence where declared. An empty optical-quality capability, such as an uninstrumented laser, is not a failed measurement.
4. When raw-pattern inspection is required, check the image-delivery receipt and its matching numerical samples. Record normalization, exposure, alignment/period and noise limitations. If vision is unavailable, report that limitation; do not claim numerical-only analysis includes visual inspection.
5. Distinguish expected sweep/target transitions, source problems, optical degradation and controller-response problems. Compare like acquisition and target regimes. A changed chart refresh alone does not establish changed feedback sampling. State plausible causes and uncertainty without inventing a confirmed physical cause.
6. Review the supplied comparable operating history: drift rate/direction, noise, time to threshold crossing after correction, intervention frequency, correction latency/success, failed or preempted adjustments and useful operating time. Separate normal locked operation from sweeps, target changes, sensor adjustments and invalid data. If history is insufficient or the context changed, state uncertainty rather than inventing a stable distribution. Use the evidence to recommend the shared operational quality criteria and, separately, correction capabilities and agent-wakeup timing. Frequent wakeups may justify more permissive task-appropriate quality criteria or different persistence/hysteresis; reliable quick recovery may justify earlier tighter correction. Do not assume faster drift always calls for relaxation, or that a stricter metric threshold necessarily improves actual performance. Choose no action, request more evidence, defer to foreground work, propose a named adjustment capability, or consult a human. Include evidence, expected benefit and required scope. Do not execute a proposed physical action or configuration change inside this diagnostic task.
7. Return the structured diagnosis and coverage record. Report a new actionable issue, material change or required human decision using the configured notification policy; keep unchanged healthy reviews in the execution record. Notification failure is separate from laser health and must not stop control.
