Command
Use list_projects({}) to confirm my supplied project host is visible. Use my chosen days value of 7, 30, or 90 for every time-windowed call. For each authorized client host, read get_overview({host, days}) and get_brand_insights({host, days}). Compare returned trends with my supplied thresholds and saved baseline export where available. Investigate exceptions using list_prompts({host, limit: 100}) and get_prompt_runs({host, prompt_id, days}). Rank cases by observed magnitude and sample coverage, separating missing data from deterioration. Draft escalation notes for review; scheduling and delivery require my independent automation connection. Return evidence links, coverage limits, and explicit unknowns; never infer unsupported scores.What you need
- Authorized client hosts, a supported window, and explicit escalation thresholds.
- A saved baseline for historical comparisons; an independent scheduler or messaging connection for recurring delivery.
Separate exceptions from missing coverage
Scan only the client projects visible to the token and approved for this review. Apply the same stated thresholds to each project, then inspect sampled answers before treating a weak signal as an issue that needs escalation.
Create a reviewable exception queue
Each exception includes the affected host, observed change, supporting prompt samples, and a suggested next check. The result is a prioritized review queue that an operator can approve before any separately configured system sends an alert or schedules another scan.