Papermorph
AGENTS.md1k starsChanged yesterdayMITView on GitHub
Papermorph is a Claude Code Skill that turns a textbook PDF, a story or an argument into a narrated, animated web book: lessons, picture books, and films (documentaries and upright shorts). This file is for an agent changing the repository; the Skill's own instructions for making books are in .claude/skills/papermorph/SKILL.md. The Skill is read by models making books, so what it says has to be true and has to be found. These rules come from what went wrong before: Write…
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 11
Don't add a copyrighted book, its PDF, or lessons made from one.
- Very specific: line 32
After an engine or page change, serve the site (python3 .claude/skills/papermorph/scripts/serve.py 8765 site), play the chapters you …
- Distinctive: line 33
After a large change to the Skill, have a fresh agent make a short film with the Skill alone, and sort what it trips on under the rules …
≈ “Common” is an estimate: at least 1 in 5 of the 36,368 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Working on Papermorph
Papermorph is a Claude Code Skill that turns a textbook PDF, a story or an argument into a narrated, animated web book: lessons, picture books, and films (documentaries and upright shorts). This file is for an agent changing the repository; the Skill's own instructions for making books are in .claude/skills/papermorph/SKILL.md.
What lives where
.claude/skills/papermorph/: the Skill.SKILL.mdroutes by tier and links every reference directly;references/core/holds what every tier shares (style, publish, sound, review, team, engine),references/lessons/,references/picture-books/andreferences/films/what is each tier's own;scripts/the tools;assets/templates/the kits a new book is copied from (scripts/new_book.py).engine/: the one source of the player and drawing engine. Change it here, then copy it into the books withpython3 .claude/skills/papermorph/scripts/engine.py site; never edit a book'slib/engine.*copy. A book's own settings live in itslib/book.css, which the sync leaves alone.site/: the published books, static files only, with no build step or backend.books/,content/andtools/are the maintainers' private working folders and are not in the repository. Don't recreate them in a commit.- Don't add a copyrighted book, its PDF, or lessons made from one.
Changing the Skill
The Skill is read by models making books, so what it says has to be true and has to be found. These rules come from what went wrong before:
- The docs describe what the kits and scripts actually do. Change the code and its docs in the same commit, and open a template before describing it. A doc once said a chapter page had map lines commented out that only another kit's page had, and an agent spent a step looking for them.
- Reference films and books are built with the kit the Skill ships. When a reference needed something the kit lacks, move it into the kit, or the reference teaches a path no one else can follow.
- Tools say what they could not do. A label pushed off frame, an unknown region name, a global that clashes with the engine: print it and name it. A silent skip looks like success to the agent running the tool.
- Shared files hold only what is shared.
references/core/is for every tier, andreferences/films/films.mdfor both documentaries and shorts. A sentence true of one tier or shape goes in that tier's file. - Each step can get what it needs from the steps before it, and each target comes with the figure to reach it. A length comes with a words-per-minute figure; a loudness comes with the level to set.
Write Skill guidance as intent, reasons and examples, and give the model ways to see its own output. Avoid lists of bans: they narrow what the model tries, and the owner has turned them down before. Keep each fact in one place and link to it.
Checks
CI (.github/workflows/checks.yml) runs these on every pull request; run them before you push:
python3 .github/scripts/check_skill.pychecks links and anchors, thatSKILL.mdlinks every reference and names every script, that scripts compile, and that flags and paths named in the docs exist.python3 .github/scripts/check_site.py(needs Node) checks that every book's engine copy matchesengine/, runslint.pyon every book, and runsnode --checkon every script, inline page scripts included.pipx run ruff check --select F,E9 --ignore F401 .claude/skills/papermorph/scripts .github enginecatches undefined names and other pyflakes errors in the Python tools.- After an engine or page change, serve the site (
python3 .claude/skills/papermorph/scripts/serve.py 8765 site), play the chapters you touched in a browser, and runpython3 .claude/skills/papermorph/scripts/lint.py site/<book>and the engine's own checks (uv run --with playwright engine/check_i18n.py,engine/check_latex.py,engine/check_find.py;engine/check_volume.pywith the site served on 8767). - After a large change to the Skill, have a fresh agent make a short film with the Skill alone, and sort what it trips on under the rules above. If a problem fits none of them, a rule is missing.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 28,693 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository2 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- papermorphSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

