agentleFS
Sign inStart free
KhaledSaeed18KhaledSaeed18/deck-kit/AGENTS.md

deck-kit

AGENTS.md0 starsChanged 12 days agoMITView on GitHub

An offline, self-hosted HTML presentation: a fixed engine (keyboard and clicker control, fragments, presenter window with sync, overview, blackout, print, single-file build, slide lint) plus one deck's content and design. Each copy of this repo is one talk.

  • 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 31,439 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 repository4 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are 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.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the host's skill picker or name the skill in your prompt.

    line 22
  2. Very specific:

    If an agent does not discover these paths, explicitly read this file and the relevant .agents/skills/<name>/SKILL.md before following the …

    line 24
  3. Very specific:

    No runtime network requests: No CDNs, remote fonts or remote images.

    line 37
  4. Very specific:

    On-screen copy voice: no emoji, no em dashes, no exclamation marks.

    line 52
  5. Distinctive:

    If delegation is unavailable, perform separate research and verification passes yourself and disclose that verification was not independent.

    line 63
  6. Distinctive:

    Commit or push only when the speaker asks.

    line 65

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

Deck kit: project rules

An offline, self-hosted HTML presentation: a fixed engine (keyboard and clicker control, fragments, presenter window with sync, overview, blackout, print, single-file build, slide lint) plus one deck's content and design. Each copy of this repo is one talk.

Read first

  • docs/brief.md: this deck's content source of truth (created by new-deck from docs/templates/brief.md). If it does not exist yet, this copy still holds the starter deck: suggest new-deck.
  • docs/decisions.md: the speaker's approved answers. Where they conflict with the brief, the decisions win.
  • docs/authoring.md: the slide format, notes structure, lint rules, icons, SVG rules, manifest.
  • docs/fact-log.md and docs/research/: verified facts. Check here before researching again.
  • examples/wifi/ and examples/crypto/: complete, finished decks. Copy their patterns (notes depth, Q&A, diagrams); examples/crypto/docs/ shows every document new-deck produces.

Agent compatibility

  • AGENTS.md is the canonical project guidance; CLAUDE.md is a relative symlink to it.
  • .agents/skills/ contains the canonical skills; .claude/skills is a relative symlink to it. Edit the canonical sources and preserve these links.
  • Use the host's skill picker or name the skill in your prompt. Codex CLI/IDE supports $new-deck; Claude Code supports /new-deck (and likewise for the other skills).
  • If an agent does not discover these paths, explicitly read this file and the relevant .agents/skills/<name>/SKILL.md before following the workflow.

Skills

  • new-deck: the whole workflow for a new talk, in phases with a review stop after each.
  • add-slide: add or rewrite one slide.
  • retheme: the design pass (touches only src/theme/** and SVG class styling).
  • ship-deck: final verification and the files to take to the room.

Hard rules

  • Never change slide content, facts or numbers from the brief on your own. Apply only approved changes from docs/decisions.md. Flag anything else, and mark unapproved on-screen text with data-todo="why" (the lint fails until it is resolved).
  • No runtime network requests. No CDNs, remote fonts or remote images. Fonts come from @fontsource* packages, icons from lucide, images live in the repo. The deck must run with Wi-Fi off.
  • Keep the three layers separate (tests enforce it):
    • src/engine/ has no visual literals and is not edited per deck. Engine changes need pnpm test:engine to pass.
    • src/slides/ holds content and semantic markup only.
    • src/theme/ holds every visual value as a token.
  • Never edit examples/: they are finished talks, and wifi/ is the engine's test fixture. Their fonts and icons are shared with your deck: do not remove a font package or icon an example still uses.
  • Stack: pnpm, Vite 8, strict TypeScript 7 (no any, enforced by Biome), vanilla DOM. No UI framework and no presentation library.
  • Icons come from lucide, hydrated from <i data-icon="name">. Never emoji or pictographic Unicode.
  • On-screen copy voice: no emoji, no em dashes, no exclamation marks. Arrows are icons. About 30 words of prose per slide (diagram labels do not count), except slides marked data-wordcount="exempt". Detail goes in the notes.
  • Diagrams are hand-authored inline SVG with dgm-* classes, a <title>, and no paint attributes. Never copy vendor diagrams or screenshots.

Working style

  • Work in phases (Brief, Research and plan, Content, Verify, Design) and stop for the speaker's review at the end of each phase.
  • Delegate research and verification to subagents when available, using the host's configured models. Use stronger reasoning for high-stakes facts and adversarial checks. If delegation is unavailable, perform separate research and verification passes yourself and disclose that verification was not independent.
  • Commit or push only when the speaker asks.
  • Report honestly: if a check fails or is skipped, say so and show the output.

Commands

  • pnpm dev (open /presenter.html alongside), pnpm build, pnpm build:single, pnpm preview
  • pnpm check (typecheck, lint, unit tests, slide lint)
  • pnpm test:deck (your deck in a browser), pnpm test:engine, pnpm test:e2e (both)
  • pnpm pdf, pnpm notes:pdf [profile], pnpm notes:md
  • pnpm dev:example and pnpm dev:example:crypto show the example decks
  • pnpm media regenerates the README images in docs/media/ (sources in docs/media/src/)

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free