agentleFS
gloom-shgloom-sh/gloomberb/AGENTS.md

gloomberb

AGENTS.md2.7k starsChanged 40 days agoMITView on GitHub

Stack: Bun + OpenTUI CONTRIBUTING.md is the contributor guide and applies to agents too: repository layout, the checks CI runs, the test policy, code rules and pull request conventions. The rules below are the short form; each links to its full text there. Tests (full policy): be selective. Add or keep a test only when it protects behavior that is easy to break and hard to catch in review: parser/math/state complexity, async/cache/persistence behavior, integration boundaries, and regressions with a concrete…

1 other project uses a version of this.

Trending

Steady talk.thin dataabout the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 0 mentionsApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 0 mentionsAug 2026: 4 mentionsSep 2026: 2 mentionsOct 2026: 1 mentionNov 2026: 1 mention
Dec 2025Jun 2026Nov 2026
Mentions a month of the gloom-sh/gloomberb repository that our weekly crawl found: only 8 in all, so this is a hint, not a trend. There is no adoption data for this file yet.

Most recently taken up by

Who's talking about gloom-sh/gloomberb

The biggest 5 of 8 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 8

Talked about on GitHub and Bluesky (8 times).

On GitHub · 7

2 more on GitHub
  • Other versions1 project uses a version of this.

    Word for word · 1

    1. vincelwt/gloomberb★ 2.7k100% 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 31,700 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

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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. Very specific:

    Use tmux to test terminal TUI changes (see the tui-testing skill).

    line 11
  2. Very specific:

    Run at most one typecheck or full bun test at a time.

    line 14
  3. Very specific:

    Never run bare gloomberb, bun start or bun dev from a non-interactive shell: the TUI never exits without a TTY, so the call hangs.

    line 16
  4. Distinctive:

    Never message, stop, archive or kill other agents or their processes, and never ask them to wait.

    line 18
  5. Distinctive:

    Never use git stash: all worktrees of a repo share one stash list, so a pop or clear can take another agent's entries.

    line 19
  6. Distinctive:

    Add mouse/cursor interactivity for everything interactive.

    line 27
  7. Distinctive:

    Never fix chart issues by disabling / turning off the kitty renderer.

    line 28

≈ “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

Stack: Bun + OpenTUI

CONTRIBUTING.md is the contributor guide and applies to agents too: repository layout, the checks CI runs, the test policy, code rules and pull request conventions. The rules below are the short form; each links to its full text there.

Tests (full policy): be selective. Add or keep a test only when it protects behavior that is easy to break and hard to catch in review: parser/math/state complexity, async/cache/persistence behavior, integration boundaries, and regressions with a concrete failure mode that could plausibly return. Static metadata, default props, pass-through wiring, copied UI text and obvious behavior are weak targets. A bug-fix test is kept only when the bug came from non-obvious behavior or a boundary likely to regress, and a test is never kept just because it exists or adds coverage. When touching a test file, trim nearby low-value tests if the cleanup is clear and low-risk.

Public text (full rule): pull request titles, descriptions, comments, commit messages and branch names are public. No internal tracker ids or links, no private repository names or links, no internal notes (deals, pricing plans, people, infrastructure); a change that needs a server change says "needs the matching server change".

Review screenshots and evidence images for a PR go through scripts/pr-assets.sh <slug> <file>..., which pushes them to refs/pr-assets/<slug> and prints the markdown links for the description: never a branch, and never committed to the tree.

Use tmux to test terminal TUI changes (see the tui-testing skill). Always kill the tmux session when done.

Shared machine: several agents often work on one host at the same time.

  • Run at most one typecheck or full bun test at a time. Each full run takes 1-3 GB, and parallel runs get OOM-killed.
  • Bound test and other risky commands with timeout -k 10 <seconds> .... A hung bun ignores the plain SIGTERM from timeout and keeps its memory for hours. Before you finish, check that nothing of yours is still running: pwdx $(pgrep -f 'bun test') 2>/dev/null | grep -F "$PWD".
  • Never run bare gloomberb, bun start or bun dev from a non-interactive shell: the TUI never exits without a TTY, so the call hangs. Use a subcommand or --help, prefix risky commands with timeout -k 10 60, and run interactive checks only in tmux.
  • Throwaway profiles (GLOOMBERB_HOME) go on disk under ~/.cache/gloom-smoke/<task>, never under /tmp (one profile is ~29k files and /tmp is a small tmpfs). Remove the profile and kill tmux when done.
  • Never message, stop, archive or kill other agents or their processes, and never ask them to wait. If the host is busy, retry later or report it.
  • Never use git stash: all worktrees of a repo share one stash list, so a pop or clear can take another agent's entries. Commit work in progress to your own branch instead.

Pane footers/status bars show changing status such as loading, error, live/delayed, stale, or auth state. Preserve existing pane-specific action shortcuts there instead of duplicating them in body toolbars. Do not add fixed pane labels, row counts, or generic keyboard hints. Never show internal provider or source identifiers (provider:*, raw feed names) anywhere in a pane; say what the data is (real-time, 15m delayed, settlement, as of) instead. Gloom Cloud is the public market-data service. See pane-conventions. Recurring methodology, model assumptions, and usage explanations belong in docs, not always-visible pane text or new info buttons. Keep units, source dates, and active data failures in context. Use existing actions instead of adding duplicate toolbars. Information density matters: never repeat the same information in a pane title/header and again in the body. If a stack/detail title already names the item, start the body with metadata or content. Desktop work, bun run typecheck and bun test need Electrobun's SDK in .hutch/devkit: run bun run desktop:sync after bun install (it also applies patches/electrobun-devkit@*.patch). For Electrobun/desktop-web-only work, do not load the OpenTUI or tui-testing skills unless the change also touches terminal OpenTUI behavior or explicitly needs tmux coverage. The OpenTUI skill is not checked in: if .agents/skills/opentui is missing, restore it with bunx skills experimental_install. For desktop/Electrobun/web UI, do not draw GUI primitives with terminal cell characters. Use real DOM/CSS/canvas/SVG primitives for lines, markers, shapes, overlays, and interaction affordances; reserve cell-character drawing for the OpenTUI terminal renderer only. Add mouse/cursor interactivity for everything interactive. Never fix chart issues by disabling / turning off the kitty renderer; preserve kitty support and fix the root cause. Built-in code under src/ imports the host by relative path (../../../ui, ../../../components, ../../../public/react, ../../../theme/colors). The gloomberb/* specifiers are the external-plugin API: Bun on Linux resolves them as a package self-reference, so typecheck and tests pass, but the Windows desktop bundle cannot resolve them and the Windows verify workflow on main fails. See Host imports. Every UI change is benchmarked across all panes and plugins, and its blast-radius sheet is attached to the PR (Holding the line): bun run check:ui-tokens must pass (UI tokens in CI: type, weight, radius and pane inset on every view, exceptions in src/architecture/ui-tokens-allowlist.json with a why), and the UI benchmark's "N of M views changed" comment and sheets (bun run ui:benchmark --base origin/main locally) show only views the change meant to move. When adding or changing a pane/plugin, read PLUGINS.md and load the pane-conventions skill (docs/pane-conventions.md); start with its Principles (one convention per element, tokens instead of raw values, extend the kit, check every UI change against all panes and plugins) and then use it for where actions, status and warnings go, detail stacks, load-more lists, tabs, forms, density, reserved keys, and the new-pane checklist. Basic UI must use the shared kit (full rule): actions, selectable/expandable rows, fields, tabs, lists/tables, headings/sections, labeled values, badges, dividers, and loading/error/empty states. Extend the kit for a missing repeated pattern and migrate its callers; do not create another pane-local implementation. Box/ScrollBox remain layout primitives; raw Text and custom surfaces are appropriate for domain content, charts, and specialized editors. External plugins use the same components through gloomberb/components.

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.