- Home
- Skills
- Agents & Orchestration
- Peek – Video Understanding for AI Coding Agents
More screenshots
Works with the AI Agents you already use
Peek – Video Understanding for AI Coding Agents
Give your AI coding agent the ability to understand videos. Extract visual frames, generate contact sheets, and create timestamped audio transcripts from video URLs or local files.
$4.99
Peek – Video Understanding for AI Coding Agents
Example session with this skill installed
- Read your context and instructions
- Compiled the peek
What it is: a 12-second vertical ad (1080×1920, 30 fps) posted by
iSTYLE Apple Premium Partner, selling the
iPhone Duo — Apple's foldable, which launched September 2026. Music only: the transcript found no speech at all.
I sampled it at 2 fps (24 stills), so anything happening between stills is inferred from the frames on either side.
Shot by shot
| Time | What's on screen |
|---|---|
| 0.0–1.5s | One hand holds the phone closed, outer screen showing the iOS home screen (San Francisco weather + map widgets). Black sans-serif "Hello." centered above. A second hand enters from bottom-left. |
| 1.5–3.5s | The device unfolds into the tablet-sized inner display. The title changes to "Hello, hello." |
| 3.5–6.5s | Apple TV app content: an editorial page titled "MATERIA", headline "RECONNECTING WITH THE LAND AT PORTUGAL'S P.A.T.E.O.S" over a sunset landscape photo. A floating player sits top-right, then goes full-screen. Tiny disclaimer at the bottom: "Apple TV requires a subscription." |
| 6.5–7.5s | The phone is set down half-open in tent/stand mode on a surface, playing SpongeBob SquarePants on the upper panel. |
| 7.5–9.5s | Held fully open in landscape, a video of a runner playing edge to edge. The dock and status icons sit vertically along the right edge — the reworked landscape layout, not the usual bottom dock. |
| 9.5–10.0s | It snaps shut, shot side-on so you see the two halves meet. |
| 10.5–12.0s | End card: Apple logo + "iPhone Duo" wordmark, hand holding the closed phone, outer screen lit. |
On-screen copy
"Hello." → "Hello, hello." → the MATERIA headline → "Apple TV requires a subscription." → "iPhone Duo"
The caption repeats Apple's spec language: largest iPhone display, thin foldable design, titanium frame, 48MP Dual Fusion camera, A20 Pro chip, "Coming soon at iSTYLE."
Caveats
No spoken audio. speech_detected: false; one non-speech segment (music) was stripped. I can't tell you anything about the track's tone or where it drops.
Motion is inferred. At 2 fps the hinge motion, the snap, and the taps are described from neighbouring stills, not watched.
- Two things I checked outside the video: the iPhone Duo is a real, shipped Apple product (announced Sept 9, 2026, on sale Oct 23), and this reel is a reseller promo rather than Apple's own ad.
Run artifacts are in C:\Users\h****\AppData\Local\Temp\opencode\peek-instagram — contact sheet, 24 frames, manifest. Say the word if you want a specific frame pulled at full size.
Connects securely to your tools. The creator never sees your data.
About this skill
Peek – Video Understanding for AI Coding Agents
Give Your AI Coding Agent Eyes for Video
AI coding agents can understand code, inspect websites, and analyze documents. But understanding the visual content of a video is a different challenge.
Peek bridges that gap.
Peek is a video-understanding skill that enables AI coding agents to inspect video content through structured visual and audio evidence.
Whether you find an interesting website animation on Instagram Reels, a UI interaction in a tutorial, or a design concept in a video, Peek helps your coding agent examine what actually happens instead of relying on a video link or your written explanation.
What Peek Does
1. Video Processing
- Accept video URLs and local video files.
- Download supported online videos using yt-dlp.
- Process video files using FFmpeg.
2. Visual Frame Extraction
- Extract representative frames from videos.
- Generate contact sheets for quick visual inspection.
- Provide structured visual evidence for your AI agent.
3. Audio Understanding
- Extract audio from video files.
- Generate timestamped transcripts using faster-whisper.
- Help agents understand spoken explanations alongside visual content.
4. Structured Evidence
- Generate manifests containing processing information and references to extracted assets.
- Organize visual and audio outputs so your agent can inspect the relevant content.
- Preserve timestamps to help identify specific moments in a video.
Why Use Peek?
Instead of manually watching a video, taking screenshots, and explaining every interaction to your AI coding agent, let Peek prepare the evidence for it.
Use Cases
- Recreating website animations and UI interactions.
- Understanding frontend design references.
- Analyzing coding tutorials.
- Inspecting product demonstrations.
- Understanding visual workflows and user interfaces.
- Turning video inspiration into actionable development context.
Requirements
- Python and the required Python packages.
- FFmpeg and ffprobe.
- yt-dlp for supported online video sources.
- faster-whisper for audio transcription.
External dependencies must be installed separately. Follow the included installation and dependency documentation.
Designed for AI Coding Workflows
Peek is designed to integrate into AI coding-agent workflows through its skill instructions and supporting tools.
It focuses on preparing video evidence for the agent to inspect, rather than automatically generating an implementation from every video.
Important Limitations
- Visual frame sampling does not capture every movement between extracted frames.
- Audio transcripts do not fully represent vocal delivery, background music, or other audio characteristics.
- Results depend on video accessibility, source compatibility, and installed dependencies.
Give your AI coding agent eyes for video. Start using Peek.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires Python 3.10+, ffmpeg, and ffprobe. Subject to a 30-minute and 1GB default limit. Instagram and private posts require a local browser for cookie access.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 12 days ago
- Software
Python 3.10+FFmpegffprobeyt-dlpfaster-whisper
faster-whisper is needed for audio transcription.
Permissions
Allowed Hosts
File Scopes
Peek uses local tools to process video files and generate extracted frames, contact sheets, audio transcripts, and structured manifests. Terminal access and file read/write permissions are required. Network access is needed when downloading videos from URLs. Local video processing does not require a cloud video-analysis API. External dependencies must be installed separately. Optional browser-cookie access may be used for supported authenticated video sources.