Inspiration
On move-out day the question is always the same: was this mark already here when I moved in? Most renters did take photos back then. They are on the NAS, mixed with thousands of other pictures, without a room name, a clause or a way to prove they have not been edited since.
What it does
The renter names the lease and two dates. The Skill reads the lease for the deposit, damage and fair-wear clauses, finds photos of each room in the move-in and move-out windows, describes visible marks without guessing cause, and saves one binder: a table per room with a move-in photo beside a move-out photo, the quoted clause, rooms with no baseline photo, and a SHA-256 value per file. Originals are only tagged, never moved.
How we built it
With the UGOS Skill Creation Tool. The parent Skill delegates to document-agent, photo-agent, file-agent and knowledge-agent; every step passes on only verified values, and every write waits for platform approval. The package passed the kit's static check with no errors or warnings.
Challenges we ran into
Keeping the Skill honest. Vision models can miss faint marks, so the binder shows photos and observations, not verdicts, and a room without a baseline photo is reported as missing instead of being filled with another room.
What we learned
The value is not in taking more photos but in finding, dating and pairing the ones that already exist.
What's next
On a finalist NAS: measure room-name recall in search-screen, check vision wording on close-ups, and confirm the approval prompts and the saved binder.
Built With
- claude
- markdown
- ugos
- uliya-skill-kit
