Overview
Johnson Agent is a safety-first Skill for UGREEN AI NAS that helps home-lab users, creators, and small teams inspect and reconcile shared media folders without relying on filename guesses or irreversible deletion.
It audits one user-confirmed NAS directory, identifies byte-identical duplicates through SHA256 checksums, flags zero-byte and unreadable files, and proposes a clearer project/ and media/ folder structure. Findings are presented for review before any filesystem changes are made.
The Problem
Shared NAS libraries often accumulate duplicate backups, inconsistent folder names, incomplete transfers, and unindexed media. A backup can appear complete while files are missing or damaged, and interrupted transfers can leave zero-byte files behind. Users may hesitate to use automated organizers because an incorrect action could move or erase important data.
The Solution
Johnson Agent combines storage-integrity diagnostics with structured media reconciliation in one Uliya Skill. It uses file metadata and cryptographic checksums to ground its findings in the actual contents of files. It does not permanently delete files. Approved duplicates are moved to quarantine, while valid files can be organized into project/ and media/ categories.
Main Workflow
- The user specifies one target directory.
- A read-only scan inspects file metadata, sizes, readability, and SHA256 checksums.
- The Skill produces a reconciliation report covering duplicates, zero-byte files, and naming or organization drift.
- The user reviews the proposed actions through the approval workflow.
- Only approved changes proceed: duplicate files move to quarantine and valid files are organized into the proposed taxonomy.
Rejecting the approval step leaves files unmoved.
Key Features
- Content-based duplicate detection: SHA256 verification identifies byte-identical files even when filenames differ.
- Integrity screening: Zero-byte and unreadable files are flagged before sorting.
- Human approval gate: Findings remain proposals until the user approves the actions through the supported interaction.
- Non-destructive quarantine: Duplicate handling avoids permanent deletion.
- Scoped execution: Each run is limited to the directory explicitly selected by the user.
- Local data handling: File metadata and checks are processed on the NAS rather than uploading the media library to an external service.
UGREEN AI NAS Integration and Architecture
The Skill follows the Uliya Skill Kit architecture. The parent Skill defines intent routing and orchestration; registered UGOS CLI experts perform filesystem inspection and approved operations.
The planned workflow uses Uliya Skill intent routing, UGOS filesystem and file-manager capabilities, and interactive approval prompts. The documented implementation references commands and capabilities including stat, sha256sum, and scan-dedupe. Read-only inspection is the default; write operations are limited to approved, quarantine-first actions.
Current Build and Validation Status
The current package includes SKILL.md and references/research-rules.md. The skill validation check passed against catalog v2.2.0 with zero reported errors. Local mock-NAS tests verified quarantine and sorting behavior.
Native execution on UGREEN UGOS hardware and multi-terabyte indexing have not yet been validated. The project is therefore presented as an early build for Round 1, not as a hardware-verified production release.
Planned Finalist Validation
- Core test 1: Byte-identical files with different names produce matching SHA256 hashes and are identified for quarantine.
- Core test 2: A flat folder containing mixed media is reconciled into project/ and media/ categories after approval.
- Edge case: A zero-byte file is flagged, and rejecting the approval prompt results in no files being moved.
Known Limitations
SHA256 hashing can be slow on multi-terabyte libraries. Execution depends on available UGOS CLI permissions and correct enforcement of the approval workflow. Permanent deletion is intentionally excluded. Kernel, firmware, and NFS replication faults are outside the current scope.
Third-Party Components and Disclosures
Built using the official UGREEN Uliya Skill Kit (v0.7.7 / Catalog v2.2.0, UGREEN Open Innovation Hackathon SDK) and standard host operating-system utilities for SHA256 hashing and filesystem operations. The proposal declares no proprietary external paid APIs or bundled third-party binaries. The Skill package's linked references should carry the complete component, source, license, and AI-tool disclosures required by the event.
Team
Syed Saad — Team Representative
Lidiya Duddekunta — Team Member
Built With
- filesystem-cli
- javascript
- linux
- sha256
- ugos
- ugreen-uliya-skill-kit