For testing: an adult character's photos, clips and portraits can use any of Venice's models (clips: any
reference-to-video one). A Venice character marked adult no longer falls back to safe for work because its photo
is Venice's: Venice's own photos count. Uploaded pictures still don't; the 18+ gate and the under-18 check stay.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BQqGFv54JJTw7p3rgGjP7C
Who owns Havre, what it is, where it stands, the decisions that hold, and the next moves. For every seat that
works on the venture, and for the board.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A row of the harbor's own above Venice's characters, and the words that go with it: how it works, and the chat's
first note.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Three tiles above Venice's characters: Maren, Odile and Taskmaster. Meeting one puts a copy of their card in your
harbor with the credit (a ported-from line, ref havre:<id>) and opens their chat; met again, it's the same copy. They
answer from their own examples without a key, so a first visit has someone to talk to.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Maren, Odile and Taskmaster live at the harbor again, as .havre.json files with a look and a voice, and as a
generated module (cast/index.js, from tools/cast.mjs) the page imports. A unit test keeps the three valid, able to
answer without a key, and the module in step with the files.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The last commit wrote ten — and one 🎉 escape as the characters themselves (the same data, but five other
ventures' lines changed). Back to json.dumps(indent=1) bytes, so the only change from before SLYNG is SLYNG's line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ww3hgdnVzTeB85eSsuPLuF
A "Copy prompt for your agent" button in the header copies SKILL.md as a prompt: one line pointing to the hosted file (for agents that can install skills) and then the skill itself, without its frontmatter, so it works in any agent, including one that can't open links. Until SKILL.md loads, or if it can't, the button copies a one-line prompt that points to it. SKILL.md teaches the format, the #d= link recipe in Python and Node, the #md= fallback for agents without code, and how to hand the link over. How it works gets a fourth card, and the README points to both. Two agents given only the copied prompt made working links, one compressed with code and one percent-encoded by hand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CYHXFvWMCs1GUmRmpySqQ6
Copy link makes a link with the deck after "#d=": its text deflated (raw) and base64url-encoded, which browsers never send to a server, so nothing is stored. The page opens a deck it finds in its address, on load or when one is pasted in later, and also takes "#md=" with the text percent-encoded, for anything that can't compress. A link is kept ready while you type so a click copies it at once, the address bar shows it after copying, and notes say when a deck came from a link, when a link is broken, and when one is long enough that some apps may cut it short. The README documents the format with Python, Node and shell snippets for agents and scripts, and its live-page link now points at mdecks.yolopreneur.lol.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CYHXFvWMCs1GUmRmpySqQ6
A slide that doesn't fit at 70% is cut into the fewest slides that do: between lines, never right after a heading, as evenly as it can, and at a heading, a blank line or a change between list and text when that costs little evenness. Each new slide repeats the title with (cont.). Cuts are measured in a hidden frame showing the copied deck at 1280x720, so the preview and the copy have the same slides on any screen. The note under the preview says which slides were cut, in grey; it turns yellow only for a line longer than a whole slide.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CYHXFvWMCs1GUmRmpySqQ6
Both the preview and the copied HTML shrink a slide's text until nothing has
to scroll, down to 70% of its size. A slide that still doesn't fit keeps
scrolling, and a note under the preview names it: "Slide 3 is too long to
fit, even with smaller text. Split it with --- or cut some words." Long words
and links now wrap instead of running off the side. Checked in Chrome: a
nine-point slide fitted at 87% in the preview, a thirty-point one was named,
and the copied deck fits the same way on a laptop and a phone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CYHXFvWMCs1GUmRmpySqQ6
agentlore/index.html still pointed its script and stylesheet at the old
path layout (/v/agentlore/assets/...), which 404s on agentlore.yolopreneur.lol,
so the page stayed blank. They now start at the root. codex-goal's link to
finishline-supervisor, the only other old /v/ path, goes to its subdomain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CYHXFvWMCs1GUmRmpySqQ6
