Inspiration and problem
Freelance editors and small creative teams need to reuse private photo libraries without losing track of which campaign shots are actually covered. BriefLens is an original UGREEN AI NAS Skill for turning a brief into an evidence-linked photo shortlist and a missing-shot report.
What the Skill does
The user supplies a typed brief or one selected document, an exact approved photo folder or album, and up to six requested shots. The workflow confirms the criteria, performs at most three scoped searches, and inspects at most 12 distinct candidate images. Each recommendation retains its returned source path, typed image identity, observed support, metadata, uncertainty and duplicate evidence. Missing shots remain missing instead of being filled with invented matches.
How it is built
The package contains original Markdown instructions using the official Uliya Skill Kit 0.7.3 and capability catalog 2.0.0. The parent delegates only to registered experts: document-agent optionally extracts a brief; photo-agent uses ug-photo search-picture, meta and vision-vlm; file-agent uses ug-filesystem stat and sha256sum for provenance and byte-duplicate evidence; knowledge-agent may save one separately approved report through ug-knowledge save-text. The parent does not directly invoke CLI or custom scripts.
Local NAS processing is preferred, with no silent external fallback. Search stays within the approved scope. Sources are not edited, moved, deleted or shared. Only completed, verified results feed dependent tasks; partial, failed, cancelled and waiting states remain explicit. Saving requires approval of the exact report and destination.
Current status and challenges
The original Skill, capability map, evidence contract and test plan were authored on October 28, 2026 with OpenAI Codex assistance. The official static checker returned zero errors and zero warnings. NAS installation, runtime execution, retrieval quality, model availability and permission enforcement remain untested. No NAS runtime success is claimed. Visual interpretation can be wrong; source rights and crop suitability require human review.
Finalist validation plan
Two core tests cover traceable shot coverage with byte duplicates, and extracting a document brief while ignoring injected instructions before verifying an approved report save. An edge test covers empty or inaccessible scope and interrupted saving.
Third-party sources and AI assistance
Official authoring reference: https://ai.ugreen.com/pages/hackathon-starter-kit . No kit code, third-party libraries, datasets or media are bundled. No independent open-source license was supplied in the kit; platform components remain subject to UGREEN terms. OpenAI Codex assisted with concept, orchestration, review and submission materials.
Track: Unleash Your Creativity. Round 1 deliverables: proposal PDF and initial Skill ZIP.
