Inspiration
A project folder can contain several files named "final" while a promised deliverable is still missing. Before handing work to a client, creators need a clear connection between the agreed brief and the files they intend to send.
HandoffLens turns that final review into an evidence-linked workflow.
What it does
HandoffLens is an AI NAS Skill designed for freelance editors, designers, and small studios.
The user selects an authorized project folder and an authoritative brief or checklist. The Skill then:
- Extracts explicit deliverables and acceptance conditions with source references.
- Inspects relevant files and available metadata within the authorized scope.
- Identifies missing outputs, technical mismatches, competing versions, and checks that remain unverified.
- Asks the creator to resolve material version choices.
- Records SHA-256 fingerprints of selected files.
- Optionally saves a reviewed Markdown report through the platform approval workflow.
Each finding connects a requirement, its source, an actual file or evidence gap, and a concrete next action.
How it is built
The initial package was developed using the official UGOS Skill Creation Tool 0.7.7 and capability catalog 2.2.0.
The Markdown Skill orchestrates three registered experts: document-agent for brief reading, file-agent for scoped file and media inspection, and knowledge-agent for approved report saving.
Dependent steps use actual returned paths, citations, metadata, and coverage information. The intended runtime uses local AI processing, with that prerequisite checked before processing private material.
Challenges addressed
A limited search cannot establish that a file is missing. A filename or timestamp cannot establish which revision the creator approved.
The workflow preserves these distinctions. It also checks for detectable file changes between inspection and fingerprinting, so earlier observations are not silently paired with a different revision.
Creative quality and client acceptance still require human review.
Current progress
The initial Skill, evidence rules, report specification, disclosures, and synthetic validation plan are complete. The package passes the official static checker with zero errors and zero warnings.
Installation and execution on a UGREEN NAS remain untested pending finalist access.
What I learned
Reliable orchestration requires clear evidence requirements, explicit scope, and honest handling of incomplete results. Successful delegation alone does not establish that the user's delivery checks are satisfied.
What's next
If selected, I will validate a complete delivery, a project with missing and conflicting outputs, and an incomplete-evidence scenario on the assigned NAS. I will verify actual metadata, fingerprints, approval handling, and saved reports, then record the working flow.
AI assistance
OpenAI ChatGPT/Codex assisted with research, ideation, naming, drafting, test design, file preparation, and static review. This assistance is disclosed in the submission materials. No NAS runtime results are claimed.
Built With
- agents
- ai
- chatgpt
- codex
- markdown
- sha-256
- ugos
- ugreen
- uliya
