CharDesk
AGENTS.md433 starsChanged 5 days agoMITView on GitHub
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.
The file
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 7
Use stable vocabulary and the fewest sufficient words.
- Distinctive: line 25
Treat the domain reference as the authority for business responsibility and the ownership map as the authority for dependency direction.
- Very specific: line 29
When shaping, implementing, or reviewing a user-facing CharDesk product interface, read DESIGN.md before loading surface-specific …
≈ “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.
Files
Docs Writing Rules
- Product instructions in
apps/docs/content/manualare written for people completing tasks. Engineering documentation assumes LLM readers know general engineering concepts. - Documentation routes knowledge to its owner; it does not reproduce tutorials or source.
- One page owns one question, and each fact has one authoritative source.
- Link authorities and connect decisions, contracts, implementation, and verification.
- Use stable vocabulary and the fewest sufficient words; remove repetition and low-value explanation.
Documentation Ownership
README.mdandREADME.zh-CN.mdare the user-facing product entry points.packages/protocol/README.mdowns protocol installation and API usage.packages/fonts/README.mdowns font installation and consumption.package.jsonowns scripts, dependencies, and local commands.apps/docs/content/manualowns the public product manual atdocs.chardesk.com; Cell UI consumer documentation belongs toui.chardesk.com.apps/docs/content/docs/developmentowns engineering contracts and Agent resources; it is excluded from product navigation and search.
Architecture Navigation
- State and data flows
- Ownership and dependency direction
- LLM–Human Text Protocol
- Domain reference
- Agent navigation benchmark
- Treat the domain reference as the authority for business responsibility and the ownership map as the authority for dependency direction; confirm implementation details in source and tests.
Product Design Routing
- When shaping, implementing, or reviewing a user-facing CharDesk product interface, read
DESIGN.mdbefore loading surface-specific authorities. - This applies to workflows, information architecture, interaction, layout, product copy, responsive behavior, accessibility, and user-visible states.
- Skip
DESIGN.mdfor backend-only work, internal protocol behavior, tests with no shipped UI impact, prose documentation, marketing pages, and content authored inside Canvas, Blackboard, CharGraph, or Slides. - For creating, editing, styling, or reviewing those artifacts, use the
chardeskskill. DESIGN.mdowns product judgment and routing. The authorities below own their implementation contracts.
Frontend Host Rules
- Visual system
- Core host icon controls
- Dialog surfaces
- Interface content
Release Automation
- User-visible commits use
fix:,feat:, orfeat!:. Release Please owns package versions,version.txt, the release manifest, changelog, tag, and GitHub release. fix:andfeat:publish a pre-1.0 patch; breaking commits publish a minor.docs:,chore:, andci:do not request a release.
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.

