Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add damionrashford/media-os --skill ffmpeg-hwaccelgit clone --depth 1 https://github.com/damionrashford/media-osWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/damionrashford/media-os/ffmpeg-hwaccel)<a href="https://agentmods.dev/skills/damionrashford/media-os/ffmpeg-hwaccel"><img src="https://agentmods.dev/badge/skills/damionrashford/media-os/ffmpeg-hwaccel/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/damionrashford/media-os/ffmpeg-hwaccel"><img src="https://agentmods.dev/badge/skills/damionrashford/media-os/ffmpeg-hwaccel.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00121 | $0.04687 |
| Opus 5 | $0.00060 | $0.02344 |
| Sonnet 5 | $0.00024 | $0.00937 |
| Haiku 4.5 | $0.00012 | $0.00469 |
Grade A, and why
ffmpeg-hwaccel scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 298 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ffmpeg Hwaccel
Context: $ARGUMENTS
Quick start
- I don't know what my GPU supports: → Step 1 (detection)
- Just encode fast on whatever GPU I have: →
scripts/hwaccel.py transcode --accel auto - Zero-copy NVIDIA pipeline (decode+scale+encode on GPU): → Step 2, recipe A
- Intel Quick Sync on Windows/Linux: → Step 2, recipe B
- VA-API on Linux desktop: → Step 2, recipe C
- macOS (Intel or Apple Silicon): → Step 2, recipe D
- AMD on Windows: → Step 2, recipe E
When to use
- You need to transcode many/long files and CPU encode is too slow.
- You want real-time or faster-than-realtime encoding (live streaming, bulk archive conversion).
- You explicitly want NVENC, QSV, VAAPI, VideoToolbox, AMF or Vulkan.
- Hardware decoder (NVDEC/QSV-dec/VAAPI-dec) to offload decoding while still using CPU for something else.
If you just want best quality at the smallest size, prefer ffmpeg-transcode with libx264/libx265/libaom-av1 — hardware encoders trade quality for speed. See the gotcha about quality-per-bit below.
Step 1 — Detect what's available on this machine
# 1. Which hwaccel APIs does this ffmpeg build expose?
ffmpeg -hide_banner -hwaccels
# 2. Which HW encoders are compiled in?
ffmpeg -hide_banner -encoders | grep -E 'nvenc|qsv|vaapi|videotoolbox|amf|vulkan'
# 3. Which HW decoders / *_cuvid variants?
ffmpeg -hide_banner -decoders | grep -E 'cuvid|qsv|vaapi|videotoolbox'
# 4. Probe the NVENC / QSV / VAAPI backends
ffmpeg -hide_banner -f lavfi -i nullsrc -c:v h264_nvenc -f null - 2>&1 | head -20 # NVENC sanity
ffmpeg -hide_banner -init_hw_device qsv=hw -f lavfi -i nullsrc -c:v h264_qsv -f null - 2>&1 | head -20
vainfo # Linux VAAPI
Or just: uv run ${CLAUDE_SKILL_DIR}/scripts/hwaccel.py detect.
Platform cheat sheet (what to pick when auto):
| OS / GPU | First choice | Fallback |
|---|---|---|
| macOS (any Mac) | videotoolbox |
libx264 |
| Linux + NVIDIA | nvenc |
vaapi |
| Linux + Intel iGPU | qsv (iHD driver) |
vaapi |
| Linux + AMD | vaapi (Mesa) |
libx264 |
| Windows + NVIDIA | nvenc |
libx264 |
| Windows + Intel iGPU | qsv |
d3d11va decode |
| Windows + AMD | amf |
libx264 |
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 298 lines · 121 tokens per session scan A 5d77dc0c3cb6
ffmpeg-hwaccel is a skill published in the GitHub repository damionrashford/media-os (18 stars, last pushed 3mo ago), licensed MIT. It adds 121 tokens to every session and 4,687 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
loudcheck
Loudness compliance verdicts against formal published standards — EBU R 128, ATSC A/85, or BS.1770 measure-only. Use it when audio or video has to pass a delivery gate: "is this file broadcast-legal?", "will this mix get rejected?", "how far off target is it and what's the exact fix?", batch-checking a delivery…
i-see-you
A cross-platform tool that takes one photograph from the computer's connected webcam and can then read, analyze, or send the image.
faceless-explainer
Turn arbitrary text — an article, notes, a topic, a brief — into a faceless explainer video: there is no site or footage to capture, so the visuals are invented per scene (typography, abstract graphics, diagrams, data-viz). Use for topic explainers, concept breakdowns, how-tos, listicles. Not a video built from a…
slideshow
Author a HyperFrames slideshow — a presentation, pitch deck, or interactive deck with discrete slides, fragment reveals, branching, hotspot navigation, and built-in presenter mode with speaker notes; also converts an existing page into a deck. Output is a navigable deck, not a rendered MP4. If the user didn't…
remotion-to-hyperframes
Port an existing Remotion (React) composition's source to HyperFrames HTML. Use ONLY on an explicit ask to port/convert/migrate/translate a Remotion source — one-way, Remotion-only. A passing Remotion mention, reference-only code, or "make something like my Remotion video" is a fresh build (/general-video). Unclear →…
hyperframes-core
The HyperFrames composition contract — build one renderable project. Use for composition structure, the data- timing attributes, class="clip", tracks, sub-compositions, variables, framework-owned media playback, deterministic-render rules, and validation. Also covers Tailwind projects and the STORYBOARD.md / SCRIPT.md…