dsh-agent-team
AGENTS.md89 starsChanged 9 days agoMITView on GitHub
This directory holds the repository's maintained engineering documentation. The root AGENTS.md keeps rules every task must know; this file only governs how to work here — routing a change to the document that owns its facts, and the discipline that keeps the set truthful. A change that invalidates a documented fact updates the owning document in the same change: The full index and Where to start paths are in [`
1 other project uses a version of this.
Adopted by
Most recently taken up by
- wowyuarm/dsh-agent-teamword for word, first seen Nov 2, 2026
Other versions1 project uses a version of this.
Word for word · 1
- wowyuarm/dsh-agent-team★ 89100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
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 31,843 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 repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .scratch/AGENTS.mdAGENTS.md
- ax-initSkill
- ax-reviewSkill
- codebase-ablationSkill
- codebase-designSkill
- diagnosing-bugsSkill
- discussionSkill
- grillingSkill
- researchSkill
- scratch-reviewSkill
- tddSkill
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 46
Write uncertain facts as > TODO: instead of guessing.
- Very specific: line 49
Commit doc changes as one Conventional Commits subject line (docs: ...) over only your own staged paths — never git add -A in the shared …
≈ “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/ Work Rules
This directory holds the repository's maintained engineering documentation. The root AGENTS.md keeps rules every task must know; this file only governs how to work here — routing a change to the document that owns its facts, and the discipline that keeps the set truthful.
Routing
A change that invalidates a documented fact updates the owning document in the same change:
| Change touches | Update |
| Domain vocabulary, concepts, or collaboration semantics | domain-model.md |
| Package ownership, the published manifest, or import direction | architecture/package-ownership.md |
| Host authority: ledger, lifecycle, projections, Session persistence, Attention, context pressure, private memory, or member capabilities | architecture/host-authority.md |
The model-facing tool set, its defining modules, or the team-member preset | architecture/tools-and-preset.md |
| Typed Remote, Client plugin and slot composition, or the Client data/presentation boundary | architecture/client-and-remote.md |
| Reuse of Harness Workspace, Session, or storage | architecture/workspace-session-storage.md |
| The collaboration model, Thread/Task/Claim semantics, or Member time awareness | team-collaboration/model-and-time.md |
| A model-facing tool contract, or a rule shared by every tool | team-collaboration/tools.md |
| Attention, Inbox, mentions, or how a Message is rendered | team-collaboration/attention-and-messaging.md |
| Mutation fences, the Human Remote boundary, or the Member context boundary | team-collaboration/boundaries.md |
| Member memory, context pressure ownership, Agent notifications, or assembled acceptance | team-collaboration/memory-and-context.md |
| Design principles or DSH design-language alignment | frontend-design/principles-and-language.md |
| Layout, typography, colour, or identity | frontend-design/layout-and-typography.md |
| A Team component or its states | frontend-design/components.md |
| Timeline scrolling, the composer, entry rows, or status pills | frontend-design/thread-and-composer.md |
| The sidebar browser or the Inbox surface | frontend-design/sidebar-browser.md |
| Refresh semantics, copy, accessibility, or evolution | frontend-design/refresh-copy-accessibility.md |
| Setup, the verification gradient, or which check a change needs | development/start-and-checks.md |
| Browser evidence, generated files, package seams, or adding a Host operation | development/generated-and-seams.md |
| Sandbox/CI environments, external installation, or release cadence | development/environments-and-install.md |
| Ledger storage routing, notification regression, or a delivery checklist | development/storage-and-delivery.md |
| Publishing a release, release gates, or release material | release-runbook.md |
DSH certification, peerDependencies, or installation checks | dsh-release-compatibility.md |
| Cross-repository lookup | harness-navigation.md |
| User-visible behavior | ../CHANGELOG.md |
The full index and Where to start paths are in README.md.
Maintenance discipline
- Source and tests define implementation behavior. When prose conflicts with code, fix the documentation — never record behavior the code does not implement.
- One fact has one maintained home; cross-link between documents instead of duplicating the fact.
- Maintained docs ship as bilingual pairs: change
foo.zh.mdin the same change asfoo.md, translating prose while keeping domain entity names in English (Agent, Member, Session, Workspace, Project, Thread, Task, Claim, preset, Remote). A Chinese rendering is correct only when it names a different concept:membershiprenders as 成员关系, but a Member entity is always Member. Interface labels keep their natural Chinese — the项目 · …chip stays as-is. - Both sides of a pair state the same rules and carry the same outline — the same headings at the same levels, in the same order, which
check:docsdecides. How a rule is rendered is the translator's call and is deliberately not compared: a list in one half and a table or prose in the other is legal. Implementation naming — a file, symbol, storage key, or CSS value — may appear on only one side. - A new maintained document gets an index row and a Where to start path in both
README.mdandREADME.zh.mdin the same change. - Write uncertain facts as
> TODO:instead of guessing. - Keep a document readable block by block: no paragraph, single list item, table cell, heading, or code line may exceed the ceiling its file records in
scripts/check-docs.mjs. Condensing a document lowers that ceiling in the same change; raising one is a deliberate decision, not a side effect. - Run
npm run check:docsafter changing a document here, adding a maintained one, or editing any of the four README pairs (the repository root and one per package) or the root contributing pair; it decides the pairing, switcher, pair-outline, relative-link, heading-anchor, index, and longest-block rules above.node scripts/check-docs.mjs --budgetsprints the per-file longest-block table instead of deciding. The shipped core skills underpackages/agent-team/core-skills/have their own gate,npm run check:core-skills. Both commands are described indevelopment/README.md. - Commit doc changes as one Conventional Commits subject line (
docs: ...) over only your own staged paths — nevergit add -Ain the shared worktree — and rungit diff --check HEADbefore committing: the bare form inspects only unstaged work, so it passes silently once your paths are staged. .scratch/is work history, not an authority; move conclusions into these documents only when durable, and never rewrite archives to match new code.
AGENTS.md files are single-language: this file and the root AGENTS.md are English, while maintained docs keep their .zh.md pairs.
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.

