What

changed

at the shopfront?

Go from a new mark to the frames around it.

Find my shopfront camera in Angelcam. Compare the live picture with the recorded frame from one hour ago. Is anything new on the facade: graffiti, broken glass or damage?

Go through the recording between 01:00 and 02:00 and find when it happened.

Every hour, compare the live picture of the Angelcam camera “Shopfront” with the recorded frame from one hour earlier, but only between 22:00 and 06:00 Prague time. Outside those hours, stop without doing anything. Ignore changes in light, rain, people walking past and parked vehicles. If there is new graffiti or damage, go through the recording of that hour, find when it happened, and keep a frame from before, during and after. Email Tomáš at tomas@yourbakery.example with what you found, the time window, the three frames and a download link for the 20 minutes around it. If glass is broken, start the subject line with “Broken glass”. Add a row to the Shopfront log at every check.

Build an artifact called Shopfront Watch. Each time I open it, read the Shopfront log sheet and show: cards for incidents in the last 30 nights, average time from damage to alert, checks run and evidence clips saved; each incident with its before, during and after frames fetched from Angelcam, the time window, what happened, the alerts sent and its status; a calendar of the last 30 nights; incidents by hour of the night; and what stands out.

Use Angelcam to find my shopfront camera. Compare the live picture with the recorded frame from one hour ago. Is anything new on the facade: graffiti, broken glass or damage?

Go through the recording between 01:00 and 02:00 and find when it happened.

Every hour, use Angelcam to compare the live picture of the camera “Shopfront” with the recorded frame from one hour earlier, but only between 22:00 and 06:00 Prague time. Outside those hours, stop without doing anything. Ignore changes in light, rain, people walking past and parked vehicles. If there is new graffiti or damage, go through the recording of that hour and find when it happened. Show a frame from before, during and after, and make a download link for the 20 minutes around it. Then notify me with what you found, the time window, and whether any glass is broken. Every check, write one log line: time, result and what you found.

Make a report for my insurer from this task’s results for the last 30 nights: incidents with the date, what happened, the time window and the evidence, how long it took from the damage to the alert, how many checks ran, and when in the night it happens. End with what stands out, and give me the incident table as a spreadsheet.

Add a custom MCP server named Angelcam at https://mcp.angelcam.com/readonly using streamable HTTP. Start its OAuth sign-in, then list the tools it offers. Do not change any cameras or recording settings.

Find my Angelcam camera “Shopfront”. Compare a current snapshot with a recorded frame from one hour earlier. Describe any new marks, broken glass or damage. Ignore lighting, rain and passing people. If the angle, visibility or image quality prevents comparison, say unclear. Do not identify people or infer a crime from appearance alone.

For the change we just reviewed, inspect recorded frames across that hour in the batches Angelcam permits. Narrow the interval between the last frame without the change and the first frame showing it. Keep before, during if available, and after frames with their timestamps. Request a recording download link around that interval if the connector supports it. If frames are missing, report the gap and do not invent an exact time.

Create a routine called Shopfront overnight review. Every hour from 22:00 through 06:00 Europe/Prague, compare a fresh snapshot of the Angelcam camera “Shopfront” with a recorded frame one hour earlier. Ignore the ordinary changes we tested. For suspected new damage, inspect the intervening recording and collect timestamped evidence. Save shopfront-log.csv with camera, check time, result, description, earliest/latest change time, the timestamps of the evidence frames and alert status. Read previous incidents so the same mark is not reported as new every hour. Post one review notification per new suspected incident in #site-team, with the camera and the timestamps of the evidence frames. If a post fails, keep the result in this chat and say why. Mark missing or incomparable images unclear. Never identify a person, make an accusation, or contact police or an insurer.

Build a page called Shopfront Watch from shopfront-log.csv. List suspected incidents with the before/after frames, observed time window, description, the camera and timestamps of the evidence frames and review status. Separate unclear checks and gaps from confirmed observations. Refresh it after each new incident. Include a reminder to download needed clips before recording retention or link validity runs out. Do not publish or share the evidence automatically.