Surveillance Watch Digest

What it does

Surveillance Watch Digest is a Uliya skill for UGREEN AI NAS that turns any chosen surveillance window into a short written digest: which cameras recorded events, who or what appeared, which people were recognised, and where the snapshot evidence lives.

What inspired us

Camera apps store events but leave the reading to you. A night with thirty motion events takes minutes of tapping, and strangers are hard to separate from familiar faces. We wanted the morning recap to read like a briefing instead of a timeline hunt.

How we built it

A single Markdown skill following the Uliya Skill Development Standard: generation-time capability evidence for the surveillance and knowledge domains, a stepwise orchestration that delegates each stage to the registered expert agents (camera list, event timeline, identity resolution, evidence export, digest assembly, optional knowledge-base save), and a references file that pins the digest format. Every step passes only the real results returned by the previous step, and empty windows, unresolved identities and refused evidence exports are handled explicitly instead of papered over.

Challenges

The hardest part was scoping the orchestration so the parent skill only delegates and never calls CLIs directly, while keeping each step's inputs strictly honest to the previous step's real results. Getting the capability declarations to match the published catalog exactly took two checker passes.

What we learned

The surveillance domain is the least crowded corner of the catalog, and identity resolution is the feature that turns a list of motion events into something a household can actually read.

Test plan

On a NAS with cameras: empty windows, single and multi camera scopes, known and stranger filtering, topic filters, unresolved identities, refused evidence exports, the optional knowledge save including the approval wait, and a word ceiling check on a dense night.

AI tools were used to draft the skill text and this story; the capability set is declared strictly against the published Uliya catalog, and the package passes the official static checker with zero errors and zero warnings.

Built With

• ai-nas
• skills-hub
• surveillance
• ugos
• ugreen
• uliya