← Back to list
video-recap Checked
A skill from the GitHub repository zenstory-ai/video-recap-skills.
9.1Overall 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-recap" skill by following the instructions at https://skill123.me/install/video-recap.
⌨️ Command line install
Run in your terminal — downloads and installs to ~/.claude/skills/.
curl -fsSL https://skill123.me/install/video-recap.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 · 110 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-recap
Score breakdown
Trigger
8.0
Rich bilingual What/When with format enumeration and trigger words (视频解说/voiceover/auto-dub/recap); no explicit When-Not boundary in the description.
Structure
9.3
216-line orchestrator SKILL.md, 6 purposeful references, 21 scripts with clean contracts; shebang present on entry point.
Workflow
9.6
Excellent orchestration design: resumable two-phase flow with run-manifest source validation, creative control modes with freeze semantics for revisions, non-negotiable audio-integrity constraints with named blockers, mandatory real-playback delivery checklist, and a --doctor environment self-check; advisory QC is deliberately fail-open.
Content
9.0
Precise, consistent writing with concrete JSON examples and full parameter enumeration; a few deep boundary rules reference sibling-skill internals.
Engineering
9.0
Valid frontmatter, entry-point shebang, env/dependency doctor, all referenced files exist; heavy env-var surface is systematically documented in config-playbook.
Security
9.5
Strong security hygiene for a cloud-API skill: all endpoints declared and overridable, keys env-only with shape-based redaction before any disk copy, QC payload explicitly excludes credentials and base64 frames; one point off because the cloud-upload nature is only visible in SKILL.md body, not the description.
Orchestrator of the zenstory-ai video-recap pipeline (understanding -> story plan -> cut -> voiceover -> assemble), with CREATE/DIRECTED/REVISION creative-control modes and resume support. Capability manifest (6.0): reads user videos, work_dir artifacts and sibling skills via the skills/ bundle root, plus API keys from env; side effects confined to work_dir, output dir, optional credential-redacted material library and optional JianYing drafts; network destinations: api.xiaomimimo.com and token-plan-{cn,sgp,ams}.xiaomimimo.com (ASR/VLM/TTS — uploads video frames, audio and narration text) and api.fish.audio (TTS incl. voice-clone reference audio), all overridable and declared in SKILL.md section 3 and config-playbook; credentials read from env only, never persisted (safe_mimo_config, redaction of key shapes in material copies); no persistence changes. Capabilities match declared scope in SKILL.md; one description-level gap: the description alone does not reveal that video content is uploaded to Xiaomi/Fish cloud. Voice-cloning dub mode is user-directed on user-supplied video (IP consideration, disclosed).