SureForge
SkillAn instruction-only skill that makes agents research before they ask, plan before they build, and verify before they call anything done.
- workflow
- verification
- quality
- skill
Investigation Video Skill
A Claude Code skill that turns one sentence into a 10-plus-minute faceless investigation video: research, script, TTS, footage, Remotion render, mixing and publishing copy.
Investigation Video Skill packages the production pipeline of a real channel that publishes two to three long-form videos a day. One prompt runs the whole chain: investigation and scriptwriting, Volcano Engine TTS narration whose per-character timestamps double as subtitle data, a per-segment footage coverage table that must be filled before rendering, Remotion compositing, FFmpeg side-chain-ducked mixing, machine acceptance checks, 3:4 and 4:3 cover generation, and copy for eight platforms validated by a script. The design rules are field-tested: every claim in the script carries a source and a date so the same ledger doubles as appeal material, six narrative types rotate so two consecutive episodes never repeat, and minimum reading-time budgets are enforced per visual element. Remotion 4.0.508 and React 19.2.8 are deliberately pinned — the author documents that upgrading broke animation timing, layout and introduced black frames.
What it does
- One prompt runs the full pipeline — investigation and script, TTS narration, footage matching, Remotion compositing, mixing, acceptance, covers and platform copy
- Subtitle timing comes straight from Volcano TTS enable_subtitle per-character timestamps at 1.2x speed — alignment is data processing, not speech recognition
- A per-segment footage coverage table gates rendering: segments without enough real moving footage get material first, never rendered with gaps
- Six narrative types (accounts, cost, rules, process, comparison, identity) rotate with distinct motion-graphic forms, and each visual element carries a minimum on-screen reading time
When to use it
Chinese-language creators running a faceless investigation or consumer-report channel who want a repeatable, machine-checked production line for 10-plus-minute horizontal videos rather than a one-off template.
Getting started
git clone https://github.com/trustfuture/investigation-video-skill.git cp -r investigation-video-skill ~/.claude/skills/ cd ~/.claude/skills/investigation-video-skill cp .env.example .env # fill in your Volcano Engine TTS keys cd template/remotion && npm install npm run still # must render one frame cd ~/.claude/skills/investigation-video-skill python3 -m pip install pillow # cover stamp + acceptance scripts # then tell Claude Code: make me a video about X
Things to know
Volcano Engine TTS credentials are required in .env — the narration step does not run without them, and the docs are Chinese throughout. The repo ships no background music: scripts/mix-bgm.sh expects your own royalty-free track at assets/bgm.mp3. Scripts under scripts/ must be copied into each episode's project directory before running (running them from the repo root fails by design). Remotion 4.0.508 and React 19.2.8 must stay pinned, and the SignPainter font needs a separate install on Windows. Node.js 18+, FFmpeg and yt-dlp via Homebrew, Python 3 plus Pillow. MIT.
Repository
trustfuture/investigation-video-skillCharted
Related on the map
Guizang's agent skill turns your repo's recent updates into a finished promo video — real components, code-rendered animation, an original code-written soundtrack — no video editor involved.
- agent-skill
- video
- promo
- marketing
geo-sleuth
SkillAn agent skill that geolocates a photo with nothing readable in it — OSM geometry, elevation skylines and satellite tiles turn “where was this taken?” into a measured answer with an error radius.
- agent-skill
- geolocation
- osint
- openstreetmap