AI Agent Hub
Back to skills
Intelligent Video Search & Retrieval Analysis icon

Intelligent Video Search & Retrieval Analysis

Design & Media Updated 2026.08.29

Paste the following prompt into your AI chat to install this skill:

Please follow https://skillhub.cn/install/skillhub.md to install @user_bb47e3e3/video-search-analysis

About this skill

Problem

Finding clips such as 'person wearing red clothes', 'talking on the phone', or a specific object in long videos is hard with keywords and manual review. This skill focuses on video search and retrieval analysis, turning target descriptions, image/video inputs, or URLs into structured retrieval tasks for security surveillance and video asset management.

How It Works

  • Three modes: conventional object retrieval, natural-language description retrieval, and vectorized model matching.
  • Inputs and outputs: supports mp4, avi, mov, local files, or network URLs; returns a structured analysis report, detection/monitoring results, suggestions, and report links.
  • Workflow: provide video input, let the system handle identity internally, run the video_search_analysis script, and review the analysis.
  • History reports: when the user asks for past reports, the skill reads the report list from the cloud API and outputs a Markdown table.

Limits and Notes

  • Search results are for reference; verify against the actual video content.
  • Use only the skill's bundled scripts instead of generating temporary scripts.
  • Network URLs are downloaded by the API service; no local pre-download is required.
  • Historical report data must come from the cloud API, not local memory or manual summaries.

Use Cases

  • Locate related surveillance clips by searching for a person wearing red clothes.
  • Search a video asset library by object or action descriptions and generate reports.
  • Search a client-provided video URL for specified targets and output detection results.
  • Query the cloud-based historical report list and format it into a table.

Best For

  • Security operations staff who need to locate people, vehicles, objects, or abnormal actions in surveillance video.
  • Video asset managers who want to retrieve clips by natural-language descriptions and keep structured reports.
  • Multimodal application engineers who need to call a video retrieval API to locate clips and parse results.
  • Data operations staff who need to review historical search report lists and verify cloud-returned fields.