claude-shorts
A Claude Code skill that guides a 10-step pipeline to extract and render vertical short-form videos from a long input video. It transcribes with word-level timestamps (faster-whisper), has Claude score and select candidate segments, snaps cut boundaries to word/sentence/audio silences, then uses Remotion to render animated caption styles and exports platform-optimized encodings (e.g., YouTube Shorts/TikTok/Instagram Reels).
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Primarily a local processing pipeline; no network API surface is documented. Security-relevant details like how prompts, transcripts, or intermediate media are stored/secured are not described. Dependency hygiene cannot be verified from README alone; GPU/FFmpeg tooling increases operational risk but not inherently security risk. TLS/auth/scoping/rate-limits are not applicable to a documented hosted API; any Claude Code authentication is handled externally.
⚡ Reliability
Best When
You want an end-to-end, local/CLI skill workflow integrated with Claude Code to go from a single input video to multiple rendered vertical shorts with consistent caption styling and boundary refinement.
Avoid When
You need a simple, stateless web API with formal guarantees about availability, latency, and idempotency; or you cannot meet the system prerequisites (FFmpeg + Node + Python and optionally NVIDIA GPU).
Use Cases
- • Turn podcasts/talks into multiple vertical clips with animated word-level captions
- • Create platform-targeted short-form exports (YouTube Shorts, TikTok, Reels)
- • Generate captioned highlights from content with improved boundary selection (word/sentence/audio-aware snapping)
- • Screen-recording shorts with cursor tracking and reframing
- • Automate an editorial workflow for selecting and rendering best-performing segments
Not For
- • A hosted SaaS API for on-demand clip generation
- • Use cases requiring fine-grained programmatic control over every intermediate artifact via HTTP endpoints
- • Environments where installing FFmpeg, CUDA/NVENC, and Node/Python dependencies is not feasible
- • Regulated workflows requiring documented compliance guarantees (not described)
Interface
Authentication
No authentication described in the README for the skill itself; interaction is via Claude Code invocation (/shorts) and local processing. Any Claude-side auth is out of scope of this repository documentation.
Pricing
Repository is MIT-licensed and appears intended for local use. Costs would mainly be compute/GPU time and Claude Code usage (not specified here).
Agent Metadata
Known Gotchas
- ⚠ Local system dependencies are required (FFmpeg, jq, bash); Windows users need WSL 2.
- ⚠ GPU acceleration is recommended but behavior with incomplete CUDA/NVENC setup is not specified in detail beyond 'falls back to CPU gracefully'.
- ⚠ The skill is interactive (segment approval and selection); a non-interactive batch mode is not described.
- ⚠ Large videos may require substantial disk space; preflight checks are mentioned but thresholds/behavior are not specified.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for claude-shorts.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.