Home / All Skills / Content & Media Creation / edit-video
English (EN)简体中文 (ZH)日本語 (JA)
Content & Media Creation · nateherkai/hyperframes-student-kit

edit-video

Edit a raw talking-head video through transcription, silence trimming, mistake review, visual storytelling, motion graphics, and verified HyperFrames rendering. Use for a complete edit; route a single requested operation directly to its specialist skill.

L4 High-Privilege · Graded auth & full auditV1 Commit-PinnedExecute CommandNetwork AccessWrite FileDual-pass Audit
Author nateherkaiGitHub nateherkai/hyperframes-student-kit ↗Stars 1070License NOASSERTION（仓库根 LICENSE，实为 MIT 文本）Commit 0d30152a82
Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

Essentially a "general manager-style" orchestration prompt: contains no scripts, relying on delegating each step to specialized skills to run the entire editing pipeline; single-point requests are routed directly to the corresponding specialist skill.

.claude/skills/edit-video/SKILL.md
Use for a complete edit; route a single requested operation directly to its specialist skill.
Only the SKILL.md file exists in the directory (verified via find), so its capability boundaries = step orchestration + disciplinary constraints + delivery contracts.

Step 1 begins with media reconnaissance and source preservation: uses ffprobe to inspect duration/streams/resolution, and either copies the source video into project assets or only references the specified local path.

.claude/skills/edit-video/SKILL.md
1. Inspect duration, streams, and source resolution with ffprobe. Copy the source into the project's assets or reference an explicitly provided local path.
Projects are uniformly placed under video-projects/<slug>/ to prevent in-place modification of raw source material.

Step 2 is the sole entry point for external services, enforcing a strict sequential order: "read cost documentation first → reuse existing transcripts → only then consider invoking services → explain uploads and costs prior to authorization."

.claude/skills/edit-video/SKILL.md
2. Read `docs/TOOLS-AND-API-KEYS.md` before service setup. Reuse a matching word transcript. Otherwise use the student's chosen provider; Nate defaults to `node scripts/transcribe-elevenlabs.mjs <source>`. OpenAI Whisper and local Whisper are supported workflow alternatives after transcript normalization. Check available credentials or local dependencies. Explain uploads and costs before any service call whose authorization is still missing.
This is the rationale for classifying this skill as Orange tier: service calls are embedded directly in its primary workflow, defaulting to the repository-level ElevenLabs helper (reads key, uploads audio, consumes quota), with options for OpenAI Whisper or local Whisper.

The editing phase is explicitly split into two hops: first generate an EDL and retimed transcript following ../cut-silences/SKILL.md, then perform review-based blooper/mistake cuts per ../cut-mistakes/SKILL.md.

.claude/skills/edit-video/SKILL.md
3. Read `../cut-silences/SKILL.md`. Produce an EDL and a retimed transcript, and render the silence pass when editing is authorized. … 4. Read `../cut-mistakes/SKILL.md`. Inspect candidates in context and preserve intentional emphasis.
Step 4 also enforces a critical discipline often overlooked: "Match every transcript to its actual video."

The motion graphics phase is likewise delegated: visual arc is handed off to video-storytelling, overlays/rhythm to hyperframes-video-beats, with output requirements to "write a beat sheet + provide one visual idea per beat + intentional callbacks."

.claude/skills/edit-video/SKILL.md
5. Read `../video-storytelling/SKILL.md` for the visual arc and `../hyperframes-video-beats/SKILL.md` for overlays. Write a beat sheet with transcript anchors, one visual idea per beat, and deliberate callbacks.
Step 6 further points to MOTION_PHILOSOPHY.md, ../hyperframes/SKILL.md, GSAP references, and style-library, requiring: "Localize every asset used by the composition."

Features two quality gates: first run preflight + HyperFrames lint + transcript sync checker and review in Studio; then, after rendering a draft, sample frames to inspect visuals and listen to cut boundaries.

.claude/skills/edit-video/SKILL.md
7. Run preflight and HyperFrames lint from the project, plus the transcript-sync validator for anchored sub-compositions. Review Studio before the draft. … 8. Render a draft, inspect encoded frames and transitions, and listen across cut boundaries.
Step 8 lists concrete inspection items: face framing / text legibility / black frames / A/V sync, requiring fixes before final rendering.

Delivery consists of a bundle of artifacts rather than a single file: final video path, edit decisions, retimed transcript, composition, and a VERIFY.md outlining "what was checked and what limitations remain."

.claude/skills/edit-video/SKILL.md
Deliver the final MP4 path, edit decisions, retimed transcript, composition, and `VERIFY.md` describing what was checked and any remaining limitations.
The final sentence of the document clarifies its capability boundary: "An automated detector proposes edits; editorial judgment remains necessary."

2Core Capabilities

01End-to-end orchestration: a unified mainline from raw talking-head footage to a verified final video (8 numbered steps)
02On-demand routing: directly triggers the corresponding specialist skill for single operations without running the full pipeline
03Transcript economy discipline: prioritizes reusing existing word-level transcripts, invoking services only when missing
04Pre-service authorization and cost disclosure (clarifies upload requirements and costs before obtaining authorization)
05Two-hop editing delegation: silence removal (EDL + retimed transcript) → blooper review editing
06Visual storytelling and rhythm orchestration delegation, requiring beat sheets and transcript anchors
07Dual quality gates: static checks (preflight/lint/transcript-sync/Studio) + post-render frame sampling and cut boundary audio inspection
08Delivery contract includes self-verifying documentation VERIFY.md (detailing checks performed and remaining limitations)

3External Dependencies

KindDependency / Endpoint
cliffprobe（媒体摸底：时长/流/分辨率）
apiElevenLabs Scribe（默认转录服务；经仓库级 helper 上传音频取词级时间戳）
apiOpenAI Whisper API（替代云转录；自带 key 与计费）
apiKie.ai（可选生成式 B-roll/图像素材服务，作者自述偏好而非必需）
cliHyperFrames CLI（preflight / lint / preview / render，路径与用法由 ../hyperframes 与 ../hyperframes-cli skill 指路）
packageNode.js + 仓库级脚本（transcribe-elevenlabs.mjs、validate-beat-sync.mjs、preflight 系列）
networkHyperFrames Studio（本地预览服务器，渲染前人工审看）

4Recommended Governance & Evidence

Suggested Controls for Safe Operation
For Agent Runtime
Graded/per-use authorization, sandbox recommended
For Administrators
Approval workflow, rollback support, full audit
Evidence & Triggers (File Paths & Excerpts)
  • scripts/: Execution of local scripts or commands: 0 scripts within this skill itself (only SKILL.md in the directory, verified via find -type f)
  • 网络端点: Declaration of external network access: Runtime: ElevenLabs transcription upload — POSTs audio to api.elevenlabs.io (via repository-level helper; SKILL.md Step 2)
  • 本地工作目录: File I/O and artifact output: Project files under video-projects/<slug>/ (source video copies, assets/, BRIEF.md or DESIGN.md, composition
Audit Notes: This skill directory contains only a single SKILL.md file (no scripts or credential-reading code). However, it defines "transcription" as Step 2 of its core workflow, defaulting to the repository-level `scripts/transcribe-elevenlabs.mjs`—a helper that reads ELEVENLABS_API_KEY and POSTs user audio to api.elevenlabs.io (sending footage off-machine and consuming account quota). It also offers OpenAI Whisper (requiring another key) and Kie.ai (bring-your-own-quota) as optional external service paths; SKILL.md explicitly mandates to "Check available credentials" and "explain uploads and costs first." According to the unified classification standard, "sending user assets to an external service backend / consuming account quota → Orange tier." This mirrors faceless-explainer (relies on HeyGen credentials for core functionality → Orange) and video-edit (drives third-party CLIs that read credentials → Orange) in the same repository. It contrasts with cut-silences / cut-mistakes (🔵), whose skill code is purely local and treats transcription as an upstream artifact (also accepting keyless local Whisper), whereas this skill codifies service invocation into its default steps and cost discipline. No Red tier characteristics were detected: no sandbox disabling, no TLS downgrades, no anti-automation bypasses, and no arbitrary code execution surfaces (rendering and CLIs run locally with parameters sourced from local files).

5Second-Pass Verification

  • [ok] 目录文件清单（是否真的无脚本） — find .claude/skills/edit-video -type f 仅返回 SKILL.md；无 scripts/、无 agents/、无隐藏文件。
  • [ok] 被引用路径是否真实存在 — 仓库内存在 .claude/skills/{cut-silences,cut-mistakes,video-storytelling,hyperframes-video-beats,hyperframes,style-library}/SKILL.md、docs/TOOLS-AND-API-KEYS.md、MOTION_PHILOSOPHY.md、scripts/transcribe-elevenlabs.mjs —— 委派链与文档依赖均非空指。
  • [ok] 转录 helper 的凭证与端点（橙档证据） — scripts/transcribe-elevenlabs.mjs 定位到：ELEVENLABS_API_KEY 三别名 env 读取 + .env 兜底、POST https://api.elevenlabs.io/v1/speech-to-text、请求头 xi-api-key、timestamps_granularity=word、ffmpeg 抽 16kHz 单声道 mp3 上传。
  • [ok] 「可选服务」是否被误判为主线 — Kie.ai 在 docs 中被标为 Optional，且明确 the kit does not bundle a universal Kie generation script；本 skill 的 SKILL.md 未直接点名 Kie，只写『学生自选 provider』——已按可选路径写入 external_deps 并在 reason 里附条件说明。
  • [ok] 自动化边界是否夸大 — SKILL.md 末句 『An automated detector proposes edits; editorial judgment remains necessary.』，与 cut-mistakes 的 review-gate 设计一致，无夸大。
  • [ok] 红档特征反查 — 全文无 sandbox 关闭、无 --no-verify/TLS 降级、无 AutomationControlled/反爬绕过、无下载并执行远端代码的指令；渲染与校验均为本地 CLI。

6Verdict & Scenarios

  • Solidifies an otherwise loose editing pipeline into an 8-step SOP, explicitly naming downstream skills and artifacts at every step with clear handoff boundaries.
  • Front-loads cost and authorization discipline: checks docs first, reuses transcripts first, and explains uploads and expenses before authorization to eliminate unnecessary cloud calls.
  • Actionable quality gates: complements static checks (preflight/lint/transcript-sync/Studio) with mandatory post-render frame sampling for faces, text legibility, black frames, and audio cut listening.
  • Includes self-verifying VERIFY.md documentation in deliverables and enforces approval deduplication (avoids re-prompting for the same authorization action).
  • Honest regarding capability boundaries: explicitly clarifies that automated detection only proposes edits while editorial judgment remains essential.
    • Best for: Best for: Creators or teams with raw talking-head footage who want to run the entire pipeline in one go—from "transcription → silence removal → mistake trimming → visual storytelling → motion graphics → rendering → verification"; also ideal as a pipeline skeleton where transcription providers or motion graphics skills can be swapped out as needed.
    Not for: Not suitable for: Users who do not want footage to leave their local machine (should switch to the local Whisper route and verify first), requests requiring only a single operation (the description directs to specialist skills directly), and environments lacking Node/ffmpeg/Chromium.
    Install Agent-Ready
    ① Verified Mirror 2026-09-28
    Option A · Direct DownloadDownload edit-video.tar.gz
    sha256: ec3a3991479999f4…
    Option B · Copy Guide for Agent
    install.json
    Agents automatically download and verify mirrors from this JSON guide.
    ② Upstream GitHub
    GitHub Repo ↗
    Pinned to commit 0d30152a82.