pitwall
AGENTS.md22 starsChanged 43 days agoMITView on GitHub
A native (Gio, pixel-drawn) terminal multiplexer for running coding agents. A daemon owns PTYs and state; the GUI client draws an aide-style sidebar with agent status next to split terminal panes. Sessions survive the window closing (daemon) and a reboot (store + agent resume). Exported signatures in each package are the contract other packages build against. Implement them as written; if one is wrong, say so in your report instead of changing it. - aide (look and behavior to match):…
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 repositoryOne other file it gives its agents.
- docs/llms.txtllms.txt
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 58
Read CONTRIBUTING.md before preparing a contribution.
- Very specific: line 66
Never list processes with arguments (ps aux, ps -ef, pgrep -a, /proc/*/cmdline) and never run bare env/printenv: the shell carries secrets.
- Very specific: line 69
Never touch ~/.claude/settings.json or ~/.codex/config.toml.
≈ “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.
Files
pitwall
A native (Gio, pixel-drawn) terminal multiplexer for running coding agents. A daemon owns PTYs and state; the GUI client draws an aide-style sidebar with agent status next to split terminal panes. Sessions survive the window closing (daemon) and a reboot (store + agent resume).
Layout
| Package | Owns |
internal/model | Shared types and aide's activity priority/tier logic |
internal/layout | Split tree types and ops |
internal/vt | Emulator contract and the emulator behind New |
internal/pane | One process in a PTY, feeding a vt.Emulator |
internal/proto | Wire messages, gob framing, socket path |
internal/daemon | State owner, socket server, wiring of pane/agent/gitstat/store |
internal/agent | Claude/Codex hook payload to model.Activity |
internal/session | Chats' events, deltas, snapshots (Reduce, Apply), Driver |
internal/guard | The deny list a guarded tab's hooks check commands against |
internal/control | ls --json rows, tab and pane addresses, the wait evaluator |
internal/progstatus | OSC 7501 (Program Status Protocol) parser and record store |
internal/gitstat | Worktrees and branch stats via git |
internal/store | Persist state, snapshots, resume commands |
internal/logind | Hold a Linux shutdown or sleep until the daemon saves |
internal/repoicon | Find a repo's root and its favicon or logo files |
internal/logs | Log files, a non-blocking log writer, a rate limiter, Redact |
internal/update | Find a newer GitHub release, install it over the binary |
internal/remote | The phone page: pairing, device tokens, its HTTP server |
internal/qr | QR codes for pairing |
internal/web | The web protocol: listener, handshake, subscriptions |
internal/input | Gio events to PTY bytes |
internal/ui/theme | Colors, derived tokens, type and spacing scales, from aide |
internal/ui/anim | Eased values, the durations, reduce_motion |
internal/ui/kit | The button, popover placement and elevation |
internal/ui/sidebar | The sidebar, and the board's drawing |
internal/ui/board | The board's columns, scopes, held cards, from the state |
internal/ui/term | Grid renderer and pane input |
internal/ui/mdview | The markdown reader: parse, flowcharts, images, its overlay |
internal/ui/app | Window, splits, Alt navigation, switcher |
cmd/pitwall | Entry point (orchestrator only) |
Exported signatures in each package are the contract other packages build against. Implement them as written; if one is wrong, say so in your report instead of changing it.
References
- aide (look and behavior to match):
../aide(a sibling checkout), especiallysrc/renderer/src/components/SidebarTree.tsx,src/shared/workspace-activity.ts,src/main/git-service.ts, the CSS tokens undersrc/renderer/src. - tuios (MIT, Go, read-only, git-excluded):
.ref/tuios. Copy whatever code we need from it. Put// Adapted from tuios (MIT): <path>above copied code. Take the part, not the framework around it.
Rules
- Read
CONTRIBUTING.mdbefore preparing a contribution. Send every change tomainthrough a pull request; GitHub blocks direct pushes for everyone. - Go 1.27.1 via
mise.toml.gofmt,go vet ./...andgo test ./...pass before every commit. - Stdlib first. A new dependency needs a reason in the commit message.
- One runnable test per non-trivial piece of logic. No test frameworks.
- GUI windows only through
agent-ws(skillagent-desktop): never launch a window on the user's desktop.agent-ws down <name>when done. - Never list processes with arguments (
ps aux,ps -ef,pgrep -a,/proc/*/cmdline) and never run bareenv/printenv: the shell carries secrets.pgrep -c -f <pattern>is fine. - Never touch
~/.claude/settings.jsonor~/.codex/config.toml.
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.

