Made with the skill it documents
Two minutes, narrated.
This explainer is itself a presentation the skill produced — authored by an agent, narrated on-device, published to a watch link. Press play.
Or open the published presentation — the exact watch link the skill printed.
What it is
A skill your agent reads natively.
No app to install, no framework to learn. The skill is a directory of instructions Claude Code, Codex, and Gemini all read natively. You describe what you want; your agent turns it into a narrated presentation and hands you a link.
How it works
Author on-agent. Narrate on-device. Host on Bisque.
Author
Your agent writes an index.html — one section per slide, the narration script in the notes of each.
Narrate
Your own machine turns the script into speech, on-device. Free and unlimited — nothing is billed by the word.
Host
Bisque packages and hosts the result at a watch URL — slides with synchronized narration, playable in any browser.
01
Check the chain.
node present.mjs doctor verifies the binary, an installed voice, your account, and a claimed handle — and synthesizes a test word to prove audio works.
02
Sign in, once.
First run walks you through browser sign-in and claims a public handle. Presentations publish under it.
03
Ask for a presentation.
“Give me a two-minute briefing on this pull request.” The agent authors the slides and narration.
04
Publish, then edit.
You get a watch link. Change a slide and publish again — only that slide is re-synthesized; a layout-only edit costs no narration at all.
Install · one command
Add the skill where your agent looks.
Any agent
npx skills add \
bisque-cloud/presenterCopies the present skill into ~/.claude/skills or ~/.agents/skills. Then ask, or invoke /present.
Manual · Claude Code
git clone https://github.com/\
bisque-cloud/presenter
cp -r presenter/skills/present \
~/.claude/skills/presentEach skill is a folder — copy the one you want where your agent looks.
Manual · Codex + Gemini
cp -r presenter/skills/present \
~/.agents/skills/presentThe shared agents location covers both. Codex ships no Node — add Node 18+ or bun.
present is the first skill in the presenter collection. Full instructions are in the README.
On your machine
Narration runs local. Free and unlimited.
Speech is synthesized on-device by bisque-voice, a ~30 MB engine your agent offers to install on first use. Nothing is billed per word — narrate as much as you like.
There's no default voice, on purpose. Your agent shows the catalogue and you pick the one that fits your disk and your taste.
- small · fastNine languages, exact word timings, ~337 MB.
- ready-madeNine studio voices, ten languages.
- your ownClone a voice from your own recording.
Requirements
What you need on the machine.
Node 18+ or bun
The skill runs on the JS runtime your CLI already uses. Codex ships none — add Node or bun alongside it.
bisque-voice
The local narration engine (~30 MB). Your agent offers to install it on first use.
A speech model
Downloaded once. There's deliberately no default — the agent shows the catalogue and names each download's size before it starts.
A Bisque account
For publishing. First run walks you through browser sign-in and claims a public handle. Narration itself needs no account.
The honest trade
Narration is local; hosting is not.
Synthesis runs on your machine for free, but publishing goes to Bisque's servers — there is no self-hosted watch page. That's the one thing the skill can't do offline, and we'd rather say so up front.
Open source, all the way down.
MIT licensed. Read the source, open an issue, send a pull request — or just clone it and ask your agent to present.