← Back to list
video-understanding Checked
A skill from the GitHub repository zenstory-ai/video-recap-skills.
8.7Overall rating
Installation
🤖 Install via AI
Copy the prompt below and send it to your AI assistant (e.g. Claude Code) — it will follow the instructions and install automatically.
Install the "video-understanding" skill by following the instructions at https://skill123.me/install/video-understanding.
⌨️ Command line install
Run in your terminal — downloads and installs to ~/.claude/skills/.
curl -fsSL https://skill123.me/install/video-understanding.sh | bash
📦 Download ZIP
Download the zip and extract it into your skills directory (e.g. ~/.claude/skills/), then restart your session.
⬇ Download v1.0.0 · 112 KBAbout this skill
Overview
A skill from the GitHub repository zenstory-ai/video-recap-skills.
Source
- Repo: https://github.com/zenstory-ai/video-recap-skills
- Path: skills/video-understanding
Score breakdown
Trigger
7.0
Clear stage description with input/output enumeration and bilingual trigger words; user-invocable=false; no explicit When-Not in the description.
Structure
9.3
Compact 75-line SKILL.md over 21 well-factored scripts and 3 references; provenance sidecars keep the cache honest.
Workflow
8.8
Solid staged pipeline with provenance-gated reuse, an observer/not-director role charter separating fact from inference, thin-evidence warnings, and bounded base64 frame caching (documented memory fix); no output self-verification loop beyond provenance matching.
Content
9.0
Precise contracts and schemas; consistent terminology across the sibling skills.
Engineering
9.0
Valid frontmatter, shebang on entry point, env requirements documented, referenced files exist.
Security
9.5
Capability manifest matches declared scope: single cloud provider (MiMo) for the core ASR/VLM function, frames/audio uploaded only for that purpose with size caps, keys env-only; one point off because the description does not state that video content is uploaded to a cloud endpoint.
Analysis stage of the zenstory-ai video-recap pipeline: scene detection, ASR, per-scene VLM observation, silence windows, fused timeline and a creative brief. Capability manifest (6.0): reads the user-supplied video file plus optional work_dir background research and env API keys; side effects confined to work_dir understanding artifacts with provenance sidecars; network destinations: api.xiaomimimo.com / token-plan clusters only — uploads extracted representative frames as base64 JPEG data URIs, audio for ASR, and optionally video scene chunks (size-capped) for the declared VLM/ASR understanding function; credentials env-only; no persistence. Capabilities match declared scope (SKILL.md section 3 is explicit about MIMO_API_KEY); minor description-level gap: cloud upload only implied via model names.