Round 1 proposal
This entry is a proposed UGREEN AI NAS Skill and an implemented Markdown package. NAS execution has not yet been validated; finalist testing would use organizer-provided access, not the entrant's personal NAS.
1. What does the Skill do?
Handover Evidence Ledger maps a work-handover checklist to cited NAS documents, exposing unsupported, contradictory, unreadable, and ownerless requirements before work changes hands.
2. Who is the target user and problem?
Small teams handing over recurring administrative work need to distinguish a checked box from evidence and avoid guessing whether a recipient accepted the work.
3. What is your proposed solution?
Use a user-approved folder and checklist to produce a requirement-by-requirement ledger, short quotations, stated owners, unresolved conflicts, and a concise handover note.
4. What is the Core User Flow?
Confirm scope and date; inventory at most twenty files; extract selected evidence; classify every checklist requirement; display the draft; save only after separate approval and read-back.
5. What substantive difference does your idea offer?
Unlike general document chat or duplicate/revision audits, the unit is a user-supplied handover obligation. A completed checkbox is not independent proof, and source support is not recipient acceptance. Comparable evidence-ledger tools exist; uniqueness is not claimed.
6. Which UGREEN AI NAS capabilities will be used, and when?
file-agent uses ug-filesystem for scoped discovery. document-agent uses ug-document extraction, questions, and reconciliation. knowledge-agent optionally uses ug-knowledge save-text after workflow approval.
7. What was your technical approach?
A Markdown orchestration Skill passes actual upstream paths and constraints to registered experts. The host NAS supplies its available AI; no external model, API key, cloud upload, or training dataset is required. Sources are read-only, citations are bounded, and missing capabilities stop execution.
8. What is the current development status?
SKILL.md and two linked references are implemented. The official checker reports static-pass with zero errors and warnings. NAS installation, extraction accuracy, approval behavior, and saving have not been validated.
9. What is your final validation plan?
Core cases: a dated source supports one requirement; conflicting sources retain both quotations. Edge case: an unreadable likely source yields unverifiable, not missing. On the assigned finalist day, verify paths, citations, denied saves, and exact-target read-back, then record English-narrated execution.
10. Are there any known limitations?
OCR and runtime capabilities may be unavailable. Document statements may be false or outdated. The workflow cannot certify completion, authenticity, compliance, or payment. Final NAS testing depends on selection and organizer access. No measured accuracy or guaranteed savings is asserted.
Third-party references
Authoring reference: official UGREEN Uliya Skill Kit and capability catalog, version 2.0.0, https://ai.ugreen.com/pages/hackathon-starter-kit . The package redistributes no kit scripts, model weights, datasets, or third-party code. Use of the organizer platform remains subject to its published terms. AI-assisted authoring was used.
