Purpose

Revision Evidence Check helps document controllers find missing or contradictory drawing revisions before circulating a document pack. Filenames and modification dates alone cannot establish which issue is current.

Proposed workflow

The user supplies one authorized NAS directory, an expected-sheet list, an authoritative issue register and the revision-order convention. A read-only Skill delegates scoped PDF discovery to file-agent and textual extraction and reconciliation to document-agent. It returns source-backed candidate revisions, coverage limits and questions for human confirmation. It does not modify files, compare CAD geometry or approve construction documents.

Development and validation

The scenario was approved by the entrant through Pavel. AI assistants prepared the proposed workflow, Markdown Skill package and synthetic decision fixtures. Official static validation and three deterministic synthetic cases were completed. These checks do not demonstrate PDF extraction, OCR, model inference or successful operation on a NAS. Those remain for the assigned finalist test environment. No real customer drawings or personal data are included.

Capabilities and limits

The design uses supported ug-filesystem find/ls and ug-document extract/qa/reconcile capabilities through registered expert agents. It is limited to thirty readable PDFs, requires actual returned paths and preserves unresolved conflicts. No paid API, external dataset or external data transfer is designed. Permissions, extraction reliability and runtime behavior need NAS validation.

Tools and provenance

Created with the official UGREEN Skill Developer Package and AI assistants. The PDF and linked disclosure identify tools, sources and development roles. No final NAS demo, repository or working-device claim is made for Round 1.

Built With

  • ugreen
Share this project:

Updates

Submission history