agentleFS
pandemicsynpandemicsyn/neondeck/AGENTS.md

neondeck

AGENTS.md8 starsChanged 2 months agoMITView on GitHub

This repository is neondeck: a local-first developer cockpit for a companion display, starting with the Corsair Xeneon Edge. The product goal is not just a chat app. Neondeck should become a small local operating system for developer context: repo registry, GitHub work queue, schedules, watches, runtime skills, memory, notifications, and durable Flue chat sessions. The roadmap is the source of truth for direction, implementation order, and completion status. Follow its current Software Factory Slice 1 status and direction and .plans/factory/SLICE_1_IMPLEMENTATION_PLAN.md…

  • Reads credentials
  • 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 32,671 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.

The file

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

9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Prefer deterministic APIs/actions for facts and mutations.

    line 38
  2. Very specific:

    Do not make the agent freestyle-edit Neondeck config as the primary path.

    line 39
  3. Distinctive:

    Keep session context stable.

    line 40
  4. Very specific:

    Treat memory as current guidance for future sessions.

    line 41
  5. Distinctive:

    Treat skills as guidance and actions as execution.

    line 44
  6. Distinctive:

    Keep UI work consistent with the existing Xeneon/Miami tiger-stripe design language unless the user asks to change direction.

    line 45
  7. Very specific:

    Add npm run changeset for user-facing package changes.

    line 112
  8. Distinctive:

    Use patch for fixes, minor for features, and major for breaking changes.

    line 113
  9. Distinctive:

    Never run prerelease or version commands in feature PRs.

    line 115

≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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

AGENTS.md

This repository is neondeck: a local-first developer cockpit for a companion display, starting with the Corsair Xeneon Edge. The product goal is not just a chat app. Neondeck should become a small local operating system for developer context: repo registry, GitHub work queue, schedules, watches, runtime skills, memory, notifications, and durable Flue chat sessions.

Read First

  • Product roadmap: .plans/ROADMAP.md
  • Hermes research notes and inspiration: .plans/HERMES_RESEARCH.md
  • User-facing project summary: README.md
  • Current agent personality file: SOUL.md

The roadmap is the source of truth for direction, implementation order, and completion status. Follow its current Software Factory Slice 1 status and direction and .plans/factory/SLICE_1_IMPLEMENTATION_PLAN.md for the requested intake-to-human-released-queue work. Build on the existing runtime, repo/GitHub, schedules, skills, and dashboard foundations; coding execution belongs to a subsequent slice. Check the roadmap and factory handoff ledger before assigning work rather than treating this direction as a completion claim.

Architecture Direction

  • Backend: Node 26, Hono, Flue.
  • Dashboard: Vite, React, Tailwind v4.
  • Marketing/docs site: Astro under docs/, deployed to Cloudflare for neondeck.dev.
  • Persistence: local SQLite. Keep Neondeck app state separate from Flue runtime state unless the roadmap changes.
  • Runtime config target: NEONDECK_HOME, then XDG_CONFIG_HOME/neondeck, then ~/.config/neondeck.

Design toward one backend command/event surface. The web dashboard is the first UI, but a future TUI/OpenTUI should reuse the same backend APIs and event streams rather than growing a second agent runtime.

Important Project Paths

  • src/: Hono/Flue backend, agents, metrics, SOUL loading, database setup.
  • web/: local dashboard SPA.
  • web/src/plugins/: typed dashboard display plugins.
  • docs/: Astro marketing/docs site.
  • config/dashboard.json: current local dashboard layout config.
  • .codex/skills/flue/: in-repo development skill for Flue.
  • .kilo/skills/flue: Kilo link to the Flue skill.
  • .plans/: planning and research documents.
  • research-repos/: local research checkouts, gitignored.

Agent Behavior Guidelines

  • Prefer deterministic APIs/actions for facts and mutations.
  • Do not make the agent freestyle-edit Neondeck config as the primary path. Config changes should go through typed Flue actions once those exist.
  • Keep session context stable. SOUL, selected skills, memory summaries, and repo config should be loaded deliberately rather than silently changing mid-session.
  • Treat memory as current guidance for future sessions. Memory uses only user, local, and project scopes. Archive stale guidance instead of deleting it so audit history remains intact.
  • Treat learning operator state as an auditable subsystem. Use /api/learning/state, the Learning dashboard tab, or neondeck learning ... for reviews, candidates, memory decisions, skill patch decisions, and audit history. Handled PR/autopilot learning events come from settled Flue agent submissions and typed local API action results with idempotent source ids. Applied skill patches may be restored only through the explicit audit-backed restore action when the target file is unchanged since application.
  • Build deterministic watchers first. Invoke agent reasoning only when there is a meaningful state change to summarize or act on.
  • Treat skills as guidance and actions as execution.
  • Keep UI work consistent with the existing Xeneon/Miami tiger-stripe design language unless the user asks to change direction.

Development Commands

Use Node 26:

fnm use 26.4.0

Fast loop:

npm run check

Individual checks:

npm run lint
npm run typecheck
npm run test
npm run test:integration
npm run test:all
npm run format:check
npm run db:check

Database migration workflow:

npm run db:generate -- --name <migration_name>
npm run db:check

Full verification:

npm run verify

Local dev:

npm run dev

Marketing/docs site:

npm run docs:astro-dev
npm run docs:build

Formatting And Linting

  • Lint: Oxlint.
  • Format: Prettier with Astro plugin.
  • Tests: Vitest.
  • npm run test runs the fast unit suite.
  • npm run test:integration runs the slower git/worktree/Kilo/autopilot workflow suites.
  • npm run test:all runs every Vitest suite.
  • npm run check is intended to stay fast and uses the unit suite.
  • npm run verify includes the full Vitest suite, format check, and full build.

Changesets

  • Add npm run changeset for user-facing package changes.
  • Use patch for fixes, minor for features, and major for breaking changes.
  • Skip changesets for docs, tests, and internal-only work.
  • Never run prerelease or version commands in feature PRs; automation owns the version PR.

Current Notes

  • .env is local-only and gitignored.
  • design/ is local reference material and gitignored.
  • research-repos/ is local research material and gitignored.
  • Build output directories are gitignored.

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.