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 novoads/agent-skills --skill broll-overlaygit clone --depth 1 https://github.com/novoads/agent-skillsWrote 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/novoads/agent-skills/broll-overlay)<a href="https://agentmods.dev/skills/novoads/agent-skills/broll-overlay"><img src="https://agentmods.dev/badge/skills/novoads/agent-skills/broll-overlay.svg" alt="Measured on agentmods" 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 | $0.00128 | $0.02401 |
| Opus 5 | $0.00064 | $0.01201 |
| Sonnet 5 | $0.00026 | $0.00480 |
| Haiku 4.5 | $0.00013 | $0.00240 |
Grade A, and why
broll-overlay 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 7d 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 — 164 lines — stays where its author put it; the contents beside it link to each section on GitHub.
B-roll overlay
Lays silent cutaway clips over a finished base video: the base audio runs untouched underneath, the picture cuts away and returns, and the final duration equals the base duration exactly. That is the whole contract — overlay, never concatenation. A previous run concatenated: 24s for a 15s base, voice dead 12s.
Assembly mechanics live entirely in scripts/broll_overlay.py; this file is
judgment only — when to enter, where the windows go, what the user sees first.
Entry condition
Two things must already exist. This skill never generates the base — if there isn't one yet, say so and stop.
-
A finished, approved base video with exactly one audio stream. Approved means the user has watched it and signed off — b-roll is a polish pass over a locked cut. The script stream-copies one voice track under the picture, so a two-stream base is rejected:
ffmpeg -i base -map 0:v -map 0:a:0 -c copy. -
Its transcript, with timings. One call to
POST /v1/transcriptswith the base'sjobId(or itsassetIdif it was uploaded). It returnstext, word-levelwords[]andsegments[]— timings in SECONDS, which is what the EDL below takes — plus ansrtyou can ignore here. Price it with the transcript arm ofPOST /v1/estimatesand quote that number, never one from this file. The meter is per minute of source, rounded up from a one-minute minimum, and a repeat of the same source is free, so asking twice costs once. No local install.(The captions endpoint still cannot supply this — it burns subtitles into a new MP4 and returns no text or timings, verified 2026-08-04 and re-checked when transcripts shipped. That finding is correct and is kept here so nobody re-discovers it and files a bug against captions.)
Offline fallback: a local whisper pass (
whisper-clioropenai-whisper) still works and is the path when there is no API key or no network. Note whisper reports timings in milliseconds where the API reports seconds, andwhisper-cliwith no model downloaded returns an EMPTY transcript rather than an error — which reads exactly like a bad render. See README for the model download.
What ships with it
4 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.
- 7d ago First seen · 164 lines · 128 tokens per session scan A 0fc0a68dd0c8
broll-overlay is a skill published in the GitHub repository novoads/agent-skills (14 stars, last pushed 12d ago), licensed MIT. It adds 128 tokens to every session and 2,401 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 2026-08-31.
Other skills, from other repositories
xiaohongshu-auto-posting
Automates the complete Xiaohongshu (XHS / Little Red Book) content operation workflow: pain-point topic collection → style case collection → topic selection → content writing → publishing → performance tracking. Use when user mentions xiaohongshu auto posting, xhs auto post, little red book posting, xiaohongshu post…
muapi-seedance-2
Expert Cinema Director skill for Seedance 2.0 (ByteDance) — high-fidelity video generation across Chinese, Global, and VIP tiers. Supports text-to-video, image-to-video, first-last-frame, omni reference, character training, omni-reference training, video editing, and watermark removal.
video-shotcraft
Create cinematic product videos from shot recipe cards, a validated template, and code/audio assets (Remotion + real page screenshots + 2.5D camera moves + beat-synced cuts + sound design). Use when the user asks to turn a frontend project or webpage into a product video, says "use video-shotcraft to make a…
muapi-social-media-video
Brand-aware social media video creator. Reads brand-identity.md, ICP.md, and messaging.md to write a post/storyboard, craft an optimized Seedance 2.0 Director prompt, generate reference frames with the best available image model, and produce platform-ready video.
muapi-ai-clipping
Turn a long video into N viral-ready short clips with a single managed API call. Wraps muapi.ai's /ai-clipping endpoint, which handles transcription, highlight ranking through a virality framework (hook / emotional peak / opinion bomb / revelation / conflict / quotable / story peak / practical value), overlap dedupe…
muapi-youtube-shorts
Auto-generate viral 9:16 YouTube Shorts (or TikTok / Reels clips) from a long-form video. Thin platform-aware wrapper around the AI Clipping skill — picks sensible defaults for short-form social platforms (9:16, 30–60s sweet spot) and delegates the actual highlight extraction + crop to muapi.ai's /ai-clipping endpoint.