agentleFS
dpalferydpalfery/kyber-weave/dash/AGENTS.md

kyber-weave / dash

AGENTS.md23 starsChanged 2 months agoMITView on GitHub

KyberDash is first-party code (ADR 0020). Judge every edit here on its merits, under the repository's own gates. Do not merge from CodeBurn or any other upstream; the only trace of it left is dash/LICENSE and dash/THIRD_PARTY_NOTICES.md. *THE TRAY AND THE WEB DASHBOARD ARE DISPLAY LAYERS. NEITHER CONTAINS ISOLATED FEATURE LOGIC, SCHEDULERS OR JOBS. ALL LOGIC AND JOBS (REFRESH SCHEDULING, CLEAN, IMPORT, PAUSE) LIVE IN THE SHARED KYBERDASH APPLICATION (THE CLI ENGINE AND THE kyberdash web SERVER BUILT FROM IT)…

  • 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 33,315 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 repository3 other files it gives its agents.

The file

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

1 rule. About 0 are common, 0 are distinctive, and 1 is 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.

  • 1

    Very specific

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

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

    Never run dash/playwright.config.ts by hand, and never omit --config.

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 42,283 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

Working under dash/

KyberDash is first-party code (ADR 0020). Judge every edit here on its merits, under the repository's own gates. Do not merge from CodeBurn or any other upstream; the only trace of it left is dash/LICENSE and dash/THIRD_PARTY_NOTICES.md.

THE RULE

THE TRAY AND THE WEB DASHBOARD ARE DISPLAY LAYERS. NEITHER CONTAINS ISOLATED FEATURE LOGIC, SCHEDULERS OR JOBS. ALL LOGIC AND JOBS (REFRESH SCHEDULING, CLEAN, IMPORT, PAUSE) LIVE IN THE SHARED KYBERDASH APPLICATION (THE CLI ENGINE AND THE kyberdash web SERVER BUILT FROM IT) THAT BOTH SURFACES SHARE. BOTH SURFACES USE ONE DATASTORE (canon.db, resolved by dash/src/canon/paths.ts resolveCanonDbPath) AND ONE API LAYER (/api/kyber/*). NO AGENT OR DESIGN MAY VIOLATE THIS.

Full text, with the boundary cases and the agent checklist: docs/rules/kyberdash-display-layer.md. Decided by ADR 0033. Architecture: docs/dash/architecture.md.

What lives where

DirectoryRole
dash/src/**The engine: parsers, the canonical store and projection, analyses, refresh and import, clean, the jobs hosts, the OTLP receiver, the server, the CLI commands
dash/web/**A display layer. Renders and requests over /api/kyber/*; holds no logic
dash/tray/**A display layer. A Tauri 2 Rust shell plus a React popover; owns pixels and lifecycle, not behaviour

Feature behaviour belongs in dash/src, reached over the API. A surface may render, invoke actions through /api/kyber/*, open links, and keep display preferences. It must never schedule, run a job, spawn a CLI child for one, host the receiver, take a non-allowlisted value import from dash/src, or compute a domain result (cost, coverage, ranking) in the browser. dash/src/architecture/display-layer.test.ts scans for exactly those and fails with the offending file, line and construct.

Gates

None of the .NET gates in the root AGENTS.md reach this tree. From the repository root, run each through the process watchdog:

scripts/proc-guard.sh run -- npm --prefix dash run typecheck
scripts/proc-guard.sh run -- npm --prefix dash run lint
scripts/proc-guard.sh run -- npm --prefix dash run test:related -- <files>
scripts/proc-guard.sh run -- npm --prefix dash run check:reachable

<files> are relative to dash/ (src/foo.ts, not dash/src/foo.ts). A repo-root path matches nothing and passes with zero tests, so check the run reports a non-zero test count.

npm --prefix dash run test is the full KyberDash suite. The review council's ts-test gate and CI run it; no dev agent does.

Editing dash/tray/** also means the tray's own gates — see dash/tray/AGENTS.md for the Rust and popover-UI commands. Run those through scripts/proc-guard.sh run -- too.

Tests that start processes

A dashboard test that starts a dashboard can start another, and that is how a run fork-bombed node in 2026-10. These rules hold for every test under dash/ (ADR 0047, test standard):

  • Playwright runs only with the isolated config, from the repository root: scripts/proc-guard.sh run --timeout 600 -- npx --prefix dash playwright test --config dash/playwright.isolated.config.ts <spec>. Never run dash/playwright.config.ts by hand, and never omit --config.
  • Never start the dashboard to run a test (kyberdash web, npm run dev -- web). The e2e boots use dash/e2e/boot-dashboard.ts, which injects inert spawners.
  • Spawns are fake by default. Inject a recording spawnFn into createNodeSpawner, DetachedLauncher or the other self-spawn seams. Under test KYBERDASH_SPAWN_GUARD=deny refuses any spawn that gets through, and the test that caused it fails.
  • A test that needs a real child calls allowRealSpawns(reason) from dash/tests/setup/spawn-guard.ts, with a reason that states the bound.
  • Every self-spawn goes through guardedSpawn in dash/src/process/spawn-guard.ts. Do not import child_process spawn for a self-spawn anywhere else; dash/tests/spawn-sites.test.ts fails the build if you do.

Adding a spawn site. A new child_process spawn, exec or fork call, a new Worker, or a Playwright webServer under dash/ fails dash/tests/spawn-sites.test.ts until it has an entry in dash/tests/spawn-sites.allowlist.ts with file, pattern, bound and note. The bound must say how the site is limited: a timeout, a tree kill, a child closed on exit, an afterEach kill, an injected fake, or a bounded external tool. A site with no limit is written "NO TIMEOUT" and listed as a follow-up; that is a residual, not a bound. Add the entry in the same change as the site. An entry covers one file and one API name, so a second site of that kind in the same file still needs its own bound stated in review.

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.