embedded-captions

An agent skill by heygen-com, from heygen-com/hyperframes. Tags: captions, design, local-workflow, office, video.

What it does

Add captions or subtitles to an existing single-subject talking-head video without editing the footage. Use for plain verbatim captions, cinematic captions embedded behind the subject, VFX captions, “炸/特效/酷炫字幕,” or a named identity from the 35-style catalog. Route by visual identity, not by backend engine. The quiet `anchor` rail is the default; embed every word only when the user explicitly wants a fully cinematic treatment. The workflow runs locally end to end, including transcription and subject matting; split multi-shot footage before applying it.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add heygen-com/hyperframes --skill embedded-captions

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/heygen-com/hyperframes
cp -r hyperframes/skills/embedded-captions ~/.claude/skills/embedded-captions

Safety box score

Grade C, moderate risk: 42 of 100 risk points, rated 2026-10-20. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill is a local video-captioning pipeline that probes footage, runs ffmpeg/Node/Puppeteer/WhisperX-style tooling, and writes rendered assets back into the project. The main risks are broad command execution, remote package/model downloads, and a project-local postprocess hook that can run arbitrary shell code; several core implementation files were not readable in this review.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	3 (high)	The workflow shells out widely (`bash`, `node`, `ffmpeg`, `npx`, `uvx`) and `scripts/render-theme.sh` will run a project-local hook: `if [[ -f "$PROJECT/_postfx.sh" ]]; then bash "$PROJECT/_postfx.sh"`.
Network access	2 (moderate)	It explicitly fetches remote content via `npx hyperframes skills update embedded-captions`, `uvx whisperx`, and first-run model/package downloads for matting and dependencies.
Filesystem reach	2 (moderate)	It writes render artifacts and helper outputs into the project (`plan.json`, `transcript.json`, `envelope.json`, HTML) and also stores model data under `~/.cache/hyperframes/background-removal/models/`.
Secrets & credentials	0 (none)	No visible code reads `.env` files, keychains, browser sessions, or other credential stores.
Privilege & persistence	1 (low)	It can modify the local environment via `npm install --prefix <project> ...` and skill updates, but there is no `sudo`, service registration, or shell/profile editing in the visible files.
Instruction hijack surface	1 (low)	The skill is highly directive (`"Never surface 'Standard vs Cinematic vs Theme' as a question"`, `--non-interactive`), but I saw no hidden prompt injection or fetched instructions being followed.
Irreversible actions	0 (none)	It renders and writes local video outputs, but I saw no posting, publishing, booking, payment, deletion, or force-push behavior.
Transparency	2 (moderate)	Several core files were not readable in this review (`scripts/make-theme.cjs`, `scripts/render-and-composite.sh`, `scripts/make-cinematic.cjs`, `scripts/check-*.cjs`, `references/composition-craft.md`, etc.), so key render/validation behavior remains unaudited.

Before installing

Read `scripts/render-theme.sh`: it can run `project/_postfx.sh` as shell code.
Confirm you are okay with networked setup: `npx hyperframes skills update ...`, `uvx whisperx`, and first-run model downloads.
Try it once in a scratch project and verify it only writes under the project plus `~/.cache/hyperframes/`.
Check the exact dependency changes from `npm install --prefix <project> --save-dev --save-exact sharp@0.35.3 puppeteer@25.8.0 gsap@3.15.0`.

Rated 2026-10-20 by an automated review of 65 files from the skill's repository.

Source

Repository: heygen-com/hyperframes (all skills from this repository)
Path: skills/embedded-captions/SKILL.md
Branch: main
Updated: 2026-09-19

captions-overlay — Overlay doctrine for the embedded-captions workflow — the caption MODEL (drop / rail / embed) and the rule that captions are an OVERLAY composited on top of.
hyperframes-creative — Non-animation creative direction for HyperFrames videos.
hyperframes-studio — Use when building or editing a HyperFrames project that people open in Studio: how the timeline should be laid out so it reads well (one caption track.
talking-head-recut — Package an existing talking-head / interview / podcast video with timed, designed GRAPHIC OVERLAY cards — kinetic titles, lower-thirds, data callouts, quotes.
ad-creative — When the user wants to generate, iterate, or scale ad creative — headlines, descriptions, primary text.
google-slides — Build and edit Google Slides through the google-slides MCP tools - slides, grids of cards and columns, text boxes, shapes, images, tables, styling.