1. What does your Skill do?
Absen checks a client's shot list against one event's photos on the NAS and tells the photographer which required shots are covered, uncertain, or still missing.
2. Who is your target user, and what is their problem?
Wedding and event photographers. Clients send a shot list: "bride with grandmother", "ring close-up", "cake cutting". Photographers cull thousands of photos, but nobody cross-checks the client's list. In Indonesian weddings with akad, resepsi and big family groups, a missed family photo cannot be reshot, and it is discovered when the client asks.
3. What is your solution?
"Absen" means roll call. During the event, after a card dump, a half-time roll call separates shots missed in finished stages from shots still coming, must-haves first, while the people are still in the room. After the event, the full roll call marks every item COVERED, WEAK, MISSING or CHECK BY EYE; run after a culling tool, it rescues required shots the culling dropped. With approval it builds a client album, tags uncertain photos, and saves the report.
4. What is the main user flow?
- "Absen the Andi-Sari list, what is still needed?"
- Absen confirms the parsed list, grouped by stage.
- The photographer points at faces: "Nenek is this unnamed person."
- Items are searched and verified.
- "Missed: #2 Bride with Nenek", taken before the reception.
5. How is your idea different from existing tools or common AI solutions?
Photo search and AI culling answer "show me X" or "which photos are good". Absen answers what should exist but does not, mid-event, when a gap can still be fixed, and hands the family helper a call sheet of who to gather. A shot with closed eyes counts as uncertain, not done. No shot-list skill exists on SkillHub today.
6. Which UGREEN AI NAS features does your Skill use, and at which step?
- Step 1: file search; comparing two list versions.
- Step 2: Photos index and face groups: is the shoot ready?
- Step 3: document extraction of the list.
- Step 4: face groups (even unnamed) and object categories.
- Step 5: search plus local vision verification.
- Steps 6–8: album, file tags, saved report, after approval.
7. What is your technical approach?
Markdown-only orchestration over four registered experts. The key judgement: trust the NAS index first and vision second. Photos already indexed faces, objects and capture times at import, so Absen filters there and runs the local vision model on usually four candidates per item, stopping at the first sufficient match. It checks readiness first, works in batches of ten, treats list text as data, and handles all five delegation states.
8. What have you built so far? What is not built yet? What still needs testing on a NAS?
Built: the complete Skill, matching rules, report format, worked example, test plan, and an automated QA suite proven by mutation tests. Nothing has run on a NAS yet. Needs a NAS: search results, vision accuracy, timings.
9. How will you test your Skill if you become a Finalist?
Staged 300-photo wedding, 20-item list, 3 items deliberately not shot.
- Core 1: those 3 MISSING; zero false COVERED.
- Core 2: half-time lists resepsi items as coming up, not missing.
- Edge: an unindexed folder stops instead of reporting false gaps.
10. What are the known limits, risks, or outside dependencies?
Main risk: half-time needs face recognition to finish between stages; I will measure minutes from card copy to recognised faces. Vision can misjudge subtle shots, so uncertain items stay WEAK. Up to 60 items per run.
Third-party components: none; no third-party license applies. Markdown only; UGOS capabilities only.
AI tools used: Claude Code for reading the developer package, drafting and reviewing the Skill Markdown, and building and running the QA checks; uliya-skill-kit 0.7.3 for catalog lookup and checks.
Built With
- claude-code
- markdown
- ugos
- uliya-skill-kit
