agentleFS
mhlagharimhlaghari/stuntman/skills/floor/SKILL.md

floor

Skill16 starsChanged 2 months agoMITView on GitHub

Open a local live board of Claude Code sessions, Codex sessions, and Stuntman workers across projects, shown as Laghari Vexel avatars. Use when the user invokes /floor, asks which agents are running or need feedback, or wants to read agent conversations and prompt ready tmux sessions from one board.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. lsnackerman/raw-clauding-skills★ 45same name
    2. per-simmons/claude-game-studio★ 4same name

    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.

  • In this repository9 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    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:

    Read host and tool setup first.

    line 8
  2. Distinctive:

    Do not add duplicate global hooks for a plugin install.

    line 46
  3. Distinctive:

    Only send the text and target the user authorized.

    line 76

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
SKILL.md

Stuntman floor

Read host and tool setup first. The floor displays lifecycle events from sessions with active hooks and runs launched through the bundled stunt wrapper. It groups agents by project and labels their host and worker status. Watching uses no model calls.

Open the board

Resolve STUNTMAN_ROOT from this installed skill, then check whether http://127.0.0.1:4517/state.json already serves the floor. If it does, open or link to http://127.0.0.1:4517/. Otherwise start the bundled helper using the host's background-process tool:

"$STUNTMAN_ROOT/bin/floor" --no-open

It prints the board URL. Open that URL with the available browser tool or give it to the user. For a terminal glance, run "$STUNTMAN_ROOT/bin/floor" --json and summarize who is working, who needs input, and who finished.

The board renders a pixel-art Dubai rooftop studio with a panorama skyline at sunset and five bundled Vexel skins mapped to hosts: Claude Code (claude), Codex (codex), Antigravity (gemini), OpenCode (deepseek), and Muse / default (laghari). Use the header controls to filter by project workstation and by agent status (Working covers running and thinking; plus Needs you, Done, Failed, Idle, Silent, Left, Other), with a Show all reset and a Showing N of M agents counter. Click a status to hide it; click again to restore it. Show all clears both project and status filters. The board is live-only with no demo mode. Audio cues for done, needs-input, and failure events are strictly opt-in via the Sound toggle or Sound Desk samples. Motion can be paused with the Motion button or via system prefers-reduced-motion. Working agents play guitar, thinking/input agents headbang with rock horns, failures rage, and completed tasks jump then celebrate.

Event setup

Plugin installs bundle the recorder in hooks/hooks.json. Start a new session after installing or updating. In Codex, the user must review and trust the hooks through /hooks before session events can arrive; never bypass trust. Do not add duplicate global hooks for a plugin install.

For the standalone helper installation, wire the intended host explicitly:

"$STUNTMAN_ROOT/bin/floor" --wire-hooks --host codex
# Or --host claude / --host both for the user's chosen hosts.

This merges settings and keeps a backup. Codex uses its hooks.json; Claude uses settings.json. Hooks only observe sessions where they are active, so an empty board does not prove no agents are running. Stuntman worker start, completion, and failure events work independently of host hook trust.

Read and send

Click an agent to read its latest conversation. The server reads public Claude/Codex transcript messages, omitting reasoning, system content, and tool results. Worker cards can show their final summary even without a transcript. Subagents are matched by their actual IDs and shown as labeled chips; separate son/daughter rigs remain a future addition.

The prompt box is enabled only for an idle or finished interactive Claude Code or Codex session in tmux. Before sending, the server verifies the recorded process identity and foreground terminal group. Busy sessions, approvals, workers, headless processes, and sessions outside tmux are view-only. Use the session's own window for approvals and questions. Resume workers through Stuntman with their original backend and returned session ID.

Sending a prompt runs work in that session and consumes its normal capacity. Only send the text and target the user authorized. The server binds to 127.0.0.1 and validates origin, host, and a per-run token for sending.

States and limits

  • Prompt submitted → thinking; tool started → running; overlapping tools stay running until all finish; turn stopped → done.
  • Permission requests and question tools → needs input. Claude's standalone Notification hook can also report permission/idle prompts.
  • Worker failure → failed; ended session → ended briefly, then removed.
  • A busy agent silent for five minutes → silent. It may still be working.
  • The event log in ~/.stuntman/floor/events.jsonl rotates near 5 MB. Writes use a short lock and fail open if the log is unavailable or contended.
  • STUNTMAN_FLOOR_DIR, STUNTMAN_FLOOR_PORT, and STUNTMAN_FLOOR_BOARD support isolated testing. Default port: 4517. --port 0 picks a free port.

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.