Claim Evidence Pack Proposal

Entrant: Nigel-Efeose Akhuetie. Track: Automate Your Workflow.

1. What it does Turns "my car was hit last night" into a sealed, hash-verified claim folder on the NAS combining camera events, phone photos and policy documents, with a summary and gap report.

2. User and problem Households with NAS cameras and phone backups facing hit-and-run, package theft, water damage or rental disputes. Evidence is scattered across cameras, photos and PDFs; insurers set notice deadlines and want specific proof, so people miss deadlines or submit weak evidence.

3. Solution One request starts a staged workflow: find what cameras saw, find corroborating photos, read the policy for deadlines and required evidence, then assemble a verified folder, stating what was not found.

4. Main user flow

1. User gives incident type and time window.
2. Camera events are listed; user confirms relevant ones.
3. Photos in the window are found; capture times checked against events.
4. Policy located; deadline, deductible, required evidence extracted with quotations.
5. One confirmation lists every file.
6. Copies go into a new claim folder (originals untouched), are hashed and verified; summary and manifest are saved. A second mode re-verifies packs.

5. Difference General assistants cannot see camera footage; cloud tools need private footage uploaded. This Skill cross-checks three local sources, writes a tamper-evident SHA-256 manifest with an anchor hash the user keeps, and reports gaps instead of inventing facts. Nothing leaves the NAS.

6. NAS features, by step Surveillance events and evidence (2); photo search, metadata, image description (3); file search (4); document extraction and Q&A (4); folder creation, copying, SHA-256 (5-6); text saving (6); local AI models and storage throughout.

7. Technical approach One SKILL.md orchestrates five registered experts (security-monitor, photo, file, document, knowledge) via delegate only. Instructions are self-contained; downstream steps use only real upstream results. Inputs: incident type, window, optional camera and policy hint. Writes follow one confirmation plus platform approvals; files are copied, never moved or deleted. Identity and plate lookups run only on request; third parties are masked in the summary. Models: NAS-local only. All five delegation states are handled.

8. Status Built: SKILL.md plus eight reference files, passing the official Uliya Skill Kit static check. Not built: runtime tuning. Pending NAS validation: real camera data, approval prompts, hash output, saving.

9. Validation Core 1: one vehicle event, one damage photo, one policy PDF; folder created, hashes match, deadline quoted. Core 2: verify mode reports an unchanged pack, then flags an added file. Edge: no events or photos; reports nothing found, offers wider search, writes nothing.

10. Limits and risks Camera evidence may be referenced, not exported. A test NAS without cameras exercises only empty-result handling. Hashes prove later integrity, not original authenticity. Not legal advice.

Third-party components and AI tools No third-party libraries, models or services; only UGOS capabilities, checked with the official Uliya Skill Kit 0.7.3. AI tools used: Grok Bot (AI assistant) drafted the Skill files and this proposal; the entrant reviewed them and is responsible.

Built With

• skill-md
• ugos
• uliya-skill-kit